Date: 2026-06-29
Author: Read-only design lane
Status: Design artifact — founder-gated, no live changes made
All facts below came from live read-only observation today.
http://127.0.0.1:3100 on the VDS (aina-vds-ts / aina-vds-tf)paperclip.service, env var PAPERCLIP_RUNTIME_API_URL=http://127.0.0.1:3100 already set/usr/bin/paperclipai, version 2026.609.07d58fc13-7c9a-4c87-b9f8-a7cfae8564a9 ("AINA")~/.hermes/profiles/aina-chief-of-staff/aina-vds-ts as user ali, cwd /srv/reposterminal(command="...") tool call (not inline-shell; skills.inline_shell: false)dispatch_in_gateway: true, 60-second poll, default_assignee: '', orchestrator_profile: ''Two dispatch paths exist. The bridge uses path A (direct, key inline) as the primary path.
paperclipai agent-prompt (top-level command)paperclipai agent-prompt \
[--api-base http://127.0.0.1:3100] \
[--title "Issue title"] \
[--no-wake] \
[--json] \
<agentRef> <agentApiKey> <prompt text...>
<agentRef> — the agent's urlKey slug (e.g. jared) or UUID. urlKey is recommended (human-readable, confirmed stable).<agentApiKey> — a per-agent bearer token issued by paperclipai token agent create --agent <id>. These are stored in the Paperclip DB, not on disk. Currently no dept-head has a pre-issued key. Keys must be provisioned and stored before the bridge can go live. See §6.<prompt text> — variadic, no quoting needed.--no-wake — creates the issue/task in Paperclip but suppresses the heartbeat wakeup signal. This is the dry-run-adjacent flag: work is filed but the agent is not triggered. Use this for the DRY-RUN TEST (§7).--api-base http://127.0.0.1:3100 — required because the server is loopback-only. Without it, the CLI would default to the cloud endpoint.--json — machine-readable output (includes issueId, agentRef, status).paperclipai agent prompt (subcommand, profile-based)paperclipai agent prompt \
--agent <agentRef> \
--api-key-env <ENV_VAR_NAME> \
[--api-base http://127.0.0.1:3100] \
[--title "Issue title"] \
[--no-wake] \
[--json] \
<prompt text...>
paperclipai context profile system exists but currently has no data — not usable yet for key storage.paperclipai agent wake \
--source on_demand \
--trigger manual \
--reason "<reason string>" \
[--payload '{"task":"..."}'] \
<agentRef>
Used after creating an issue with --no-wake to trigger the agent separately (e.g., once a confirmation step passes).
The 12 direct reports to Jessica (CEO, urlKey jessica, UUID 6454b8e0). These are the valid dispatch targets for the bridge.
| Dispatch Name | urlKey | UUID | Domain |
|---|---|---|---|
| Laurie | laurie |
af273e31-0f40-4a91-b635-09a11534f5aa |
Data & Personalization |
| Monica | monica |
379acc14-2207-432d-92a7-25ab0a85c1fa |
Product & Learning |
| Mike | mike |
eb3f53b4-a347-4e19-9de2-593830c00d54 |
Research & Intelligence |
| Chief Compliance Officer | chief-compliance-officer |
5fb74c15-bcb2-464a-ba2c-199d6f2b5b1c |
Compliance (paused) |
| Frodo | frodo |
88b49386-ac86-4b3f-876f-4a0af410fdfe |
Release Management |
| Erlich | erlich |
773887c3-05b3-496a-ab5c-7ca1ff417233 |
Media |
| Harvey | harvey |
85254289-bcfa-4c81-b001-706620004ca4 |
Growth |
| Benjamin | benjamin |
51b3bd27-a37f-4901-82f8-c1c02f63ce19 |
Security & Privacy |
| Jared | jared |
a873590c-2166-4054-9dad-329e17d2c889 |
AgentOps |
| Richard | richard |
be6cc169-b8b4-4d96-9749-825d63aceecf |
Platform / CTO |
| Gandalf | gandalf |
a40a26d9-8a45-483e-9232-f46b82e5c382 |
Chief of Staff (peer) |
| Moneypenny | moneypenny |
b5e5c393-cd14-461a-b771-8958b60fb001 |
Concierge |
Note: chief-compliance-officer is currently paused. The bridge must refuse to dispatch to paused agents. Jessica is NOT a dispatch target — she is the orchestrating CEO, not a task recipient for external callers.
Path: ~/.hermes/profiles/aina-chief-of-staff/skills/dispatch-to-paperclip/SKILL.md
Use the CoS short-form frontmatter (matching delegate-to-specialist, file-runfusion-mission, morning-brief). This is consistent with the profile's existing custom-skill pattern.
Content shape:
---
name: dispatch-to-paperclip
description: Use when Donna needs to assign a task directly to a Paperclip department-head agent — filing the task into their issue queue and optionally triggering their wakeup. Bridges the Hermes kanban/coordination plane to the Paperclip execution plane.
when_to_use: Ali or the kanban board routes a task to a Paperclip dept head (Laurie, Monica, Mike, Erlich, Harvey, Benjamin, Jared, Richard, Gandalf, Moneypenny, Frodo) rather than a Hermes worker profile or RunFusion mission.
---
## Purpose
Issue a `paperclipai agent-prompt` command on the VDS to file a task into a Paperclip dept-head agent's issue queue. The CoS is the router/coordinator; the Paperclip agent is the executor.
## Trigger
A task is classified for a Paperclip dept head AND:
- `PAPERCLIP_DISPATCH_ENABLED=true` is set in the VDS environment, AND
- the target agent status is `idle` (never dispatch to `paused` or `active` unless Ali explicitly overrides)
## Pre-dispatch checks (run before issuing the command)
1. Confirm `PAPERCLIP_DISPATCH_ENABLED` is `true` — if absent or `false`, log "bridge disabled, not dispatching" and stop.
2. Confirm the target agent is not paused: `paperclipai agent list --company-id 7d58fc13-7c9a-4c87-b9f8-a7cfae8564a9 --json` → check `.status` for the target urlKey.
3. Confirm the agent API key env var is set: `echo ${PAPERCLIP_KEY_<URLKEY_UPPER>:-unset}` — if `unset`, stop and report which key is missing.
4. In DRY-RUN mode (`PAPERCLIP_DISPATCH_DRY_RUN=true`): print the exact command that WOULD be issued, then add `--no-wake`, run it, log the resulting issue ID, then immediately comment "DRY-RUN: issue created but agent not woken" on that issue.
## Routing table
| Domain / intent | Target urlKey |
|---|---|
| Data, personalization, calibration, embeddings | laurie |
| Product, learning, curriculum, mobile | monica |
| Research, intel, evidence, synthesis | mike |
| Media, audio, TTS, editorial | erlich |
| Growth, landing, SEO, social | harvey |
| Security, privacy, compliance, secrets | benjamin |
| AgentOps, incident, routines, VDS ops | jared |
| Platform, infra, code, API, auth, payments | richard |
| Coordination, memory, canon steward | gandalf |
| Run-mode concierge, intake | moneypenny |
| Release management | frodo |
| Compliance (PAUSED — do not dispatch) | chief-compliance-officer |
## Command template (live mode)
```bash
terminal(command="""
PAPERCLIP_KEY=$(printenv PAPERCLIP_KEY_<URLKEY_UPPER>)
paperclipai agent-prompt \
--api-base http://127.0.0.1:3100 \
--title "<task title, max 80 chars>" \
--json \
<urlKey> "${PAPERCLIP_KEY}" \
"<one-paragraph task spec with acceptance criteria>"
""")
terminal(command="""
PAPERCLIP_KEY=$(printenv PAPERCLIP_KEY_<URLKEY_UPPER>)
paperclipai agent-prompt \
--api-base http://127.0.0.1:3100 \
--title "[DRY-RUN] <task title>" \
--no-wake \
--json \
<urlKey> "${PAPERCLIP_KEY}" \
"<task spec>"
""")
issueId; log "Dispatched to <urlKey>, issueId=<id>"issueId; log "DRY-RUN issue <id> created, agent NOT woken"chief-compliance-officer (paused)jessica (she is the CEO root, not a task recipient)PAPERCLIP_DISPATCH_ENABLED first
---
### 4B. VDS Environment File (API keys + activation switch)
**Path on VDS:** `/home/ali/.paperclip/instances/default/bridge.env`
This file is sourced by the bridge before any dispatch. It stores the activation switch and per-agent API keys.
**Content shape:**
```bash
# Hermes→Paperclip bridge configuration
# ACTIVATION SWITCH — defaults OFF; set to "true" to enable live dispatch
PAPERCLIP_DISPATCH_ENABLED=false
# DRY-RUN mode — when enabled, files issues with --no-wake (no agent triggered)
# Set to false only after dry-run testing passes
PAPERCLIP_DISPATCH_DRY_RUN=true
# API base (loopback, do not change)
PAPERCLIP_API_BASE=http://127.0.0.1:3100
# Per-dept-head agent API keys
# Generate with: paperclipai token agent create --agent <uuid> --company-id 7d58fc13...
# Then paste the token value below (no quotes needed)
PAPERCLIP_KEY_LAURIE=
PAPERCLIP_KEY_MONICA=
PAPERCLIP_KEY_MIKE=
PAPERCLIP_KEY_ERLICH=
PAPERCLIP_KEY_HARVEY=
PAPERCLIP_KEY_BENJAMIN=
PAPERCLIP_KEY_JARED=
PAPERCLIP_KEY_RICHARD=
PAPERCLIP_KEY_GANDALF=
PAPERCLIP_KEY_MONEYPENNY=
PAPERCLIP_KEY_FRODO=
# NOTE: chief-compliance-officer is paused — no key provisioned
The SKILL.md references PAPERCLIP_KEY_<URLKEY_UPPER> — the convention is urlKey uppercased with hyphens replaced by underscores (e.g., chief-compliance-officer → PAPERCLIP_KEY_CHIEF_COMPLIANCE_OFFICER, but since that agent is paused it has no key).
The Hermes terminal tool runs commands via SSH to aina-vds-ts. The dispatch command should source this file:
terminal(command="source /home/ali/.paperclip/instances/default/bridge.env && ...")
Today the kanban dispatcher routes tasks to Hermes worker profiles only. There is no Paperclip-aware routing built in. Two adapter options exist:
Option 1 — Hermes worker profile as a relay (recommended, lower risk)
Create a thin Hermes worker profile named paperclip-dispatcher whose only behavior is: pick up the kanban task, extract the assignee_dept_head field from the task metadata, call the dispatch-to-paperclip skill, and mark the kanban task done when the issue is filed.
Config change required in ~/.hermes/profiles/aina-chief-of-staff/config.yaml:
# No change to the kanban section itself.
# A kanban task created with assignee="paperclip-dispatcher" will be picked up
# by the gateway and spawn the paperclip-dispatcher profile.
# The paperclip-dispatcher profile must exist at:
# ~/.hermes/profiles/paperclip-dispatcher/
Create ~/.hermes/profiles/paperclip-dispatcher/config.yaml (minimal):
name: paperclip-dispatcher
toolsets:
- hermes-cli
terminal:
backend: ssh
ssh_host: aina-vds-ts
ssh_user: ali
skills_dirs:
- ~/.hermes/profiles/aina-chief-of-staff/skills/dispatch-to-paperclip
This profile inherits the dispatch-to-paperclip skill and is invoked by the gateway whenever Donna creates a kanban task with assignee="paperclip-dispatcher".
Option 2 — Direct field in kanban task (no new profile)
Donna creates the kanban task with a custom metadata field:
kanban_create(
title="<task title>",
description="<spec>",
assignee="paperclip-dispatcher", # routed to the relay profile
metadata={
"paperclip_target_urlkey": "jared",
"paperclip_task_type": "agentops"
}
)
The paperclip-dispatcher profile reads HERMES_KANBAN_TASK (auto-injected env var), extracts metadata.paperclip_target_urlkey, and calls the dispatch-to-paperclip skill with that target.
Recommendation: Start without kanban wiring. Donna calls dispatch-to-paperclip directly from conversation — no kanban relay needed for the first integration. Add the paperclip-dispatcher profile in a follow-up once direct dispatch is proven.
Ali (or kanban card) → Donna (Hermes, aina-chief-of-staff profile)
│
├─ Donna classifies the task domain → picks target urlKey from routing table
│
├─ Pre-checks (via terminal tool over SSH to aina-vds-ts):
│ ├─ source /home/ali/.paperclip/instances/default/bridge.env
│ ├─ Check PAPERCLIP_DISPATCH_ENABLED=true
│ ├─ Check PAPERCLIP_DISPATCH_DRY_RUN=false (or respect dry-run)
│ ├─ paperclipai agent list --company-id 7d58fc13... --json → confirm target is idle
│ └─ printenv PAPERCLIP_KEY_<URLKEY_UPPER> → confirm key is set
│
├─ Issue the command (live mode):
│ paperclipai agent-prompt \
│ --api-base http://127.0.0.1:3100 \
│ --title "<task title>" \
│ --json \
│ <urlKey> "${PAPERCLIP_KEY_...}" "<task spec + acceptance>"
│
├─ Output: JSON → extract issueId
│ Log: "Dispatched <task> to <urlKey>, issueId=<id>"
│
└─ Paperclip server receives the prompt, creates an issue in the agent's queue,
sends heartbeat wakeup to the agent's Codex adapter
│
└─ Agent (Codex local, gpt-5.5) wakes, reads AGENTS.md, processes the issue
Output JSONL: /home/ali/.codex/sessions/<date>/<session-id>.jsonl
(same path as all other Codex sessions; no separate output location)
Output capture: The agent's work appears as a normal Codex session JSONL in /home/ali/.codex/sessions/. Donna can monitor by polling paperclipai agent list --company-id ... --json to watch for the target agent's status to return to idle, then use paperclipai issue list --agent <urlKey> (or equivalent) to fetch the completed issue body.
Currently no dept-head has a pre-issued API key. Keys live in the Paperclip DB; the only way to get one is:
# Run on VDS (via vds-exec.sh wrapper):
paperclipai token agent create \
--agent <agent-uuid> \
--company-id 7d58fc13-7c9a-4c87-b9f8-a7cfae8564a9
This prints the token once at creation time. It must be captured immediately and written into bridge.env under the corresponding PAPERCLIP_KEY_<URLKEY_UPPER> variable.
Order of operations (one-time setup, founder-gated):
1. Provision keys for the agents you want to dispatch to (start with jared for AgentOps as the safest test target).
2. Write each key into bridge.env.
3. Run the DRY-RUN TEST (§7).
4. Flip PAPERCLIP_DISPATCH_ENABLED=true and PAPERCLIP_DISPATCH_DRY_RUN=false only after dry-run passes.
Key security: The bridge.env file is at /home/ali/.paperclip/instances/default/bridge.env — inside the already-existing Paperclip secrets directory, permission 600. Never log the key value; always redact in Donna's output.
This test proves the bridge picks the right dept head and would issue the correct command without waking any agent.
bridge.env has at least one API key set (e.g., PAPERCLIP_KEY_JARED)PAPERCLIP_DISPATCH_ENABLED=true (needed to get past the enable check)PAPERCLIP_DISPATCH_DRY_RUN=true (suppresses wakeup)Ask Donna: "File a test task to the AgentOps dept head: 'Confirm VDS disk utilization is below 80% and report back.'"
jaredbash
source /home/ali/.paperclip/instances/default/bridge.env && \
paperclipai agent-prompt \
--api-base http://127.0.0.1:3100 \
--title "[DRY-RUN] VDS disk utilization check" \
--no-wake \
--json \
jared "${PAPERCLIP_KEY_JARED}" \
"Confirm VDS disk utilization is below 80% and report back."issueId<id> created for jared, agent NOT woken"jared — not another agent. Correct.--no-wake flag is present. Correct.idle (no heartbeat sent). Verify: paperclipai agent list --company-id 7d58fc13... --json | jq '.[] | select(.urlKey=="jared") | .status'paperclipai issue list --agent jared --json | jq '.[0].title' → should show the DRY-RUN title./home/ali/.codex/sessions/ for Jared during the test window).File: /home/ali/.paperclip/instances/default/bridge.env
Variable: PAPERCLIP_DISPATCH_ENABLED
| Value | Effect |
|---|---|
false (default) |
Bridge is completely inert; dispatch-to-paperclip skill logs "bridge disabled" and returns without issuing any command |
true (founder must set explicitly) |
Bridge attempts dispatch (subject to dry-run flag) |
Secondary switch: PAPERCLIP_DISPATCH_DRY_RUN
| Value | Effect |
|---|---|
true (default) |
Issues with --no-wake; agent is not triggered |
false (founder must set explicitly) |
Full live dispatch with wakeup |
Activation sequence:
1. Founder provisions API keys
2. Sets PAPERCLIP_DISPATCH_ENABLED=true (dry-run stays true)
3. Runs DRY-RUN TEST
4. Sets PAPERCLIP_DISPATCH_DRY_RUN=false for live dispatch
| Item | Reason |
|---|---|
API key provisioning (paperclipai token agent create) |
Issuing keys grants bearer-token access to wake and task agents. Irreversible until the key is rotated. |
Setting PAPERCLIP_DISPATCH_ENABLED=true |
First time live commands reach the Paperclip API. |
Setting PAPERCLIP_DISPATCH_DRY_RUN=false |
First time agents are actually woken by external dispatch. |
Adding kanban routing (paperclip-dispatcher profile) |
Automates dispatch without per-task Donna review; wider blast radius. |
No static keys today. The primary unblocking prerequisite is key provisioning. Until keys exist in bridge.env, the bridge is inert regardless of the enable flag.
chief-compliance-officer is paused. The skill must hard-block dispatch to that agent. Silently routing to a paused agent results in a task that sits forever.
API base is loopback. --api-base http://127.0.0.1:3100 is required on every command. Without it, the CLI hits the cloud endpoint and will fail (401 or wrong company). This must be in every terminal command the skill issues.
Key leakage via logs. Donna's terminal output must be redacted before being surfaced to Ali or written to any PKM artifact. The skill SKILL.md explicitly says "do not log the key."
Mass-paused subagents (2026-06-26 event). Most leaf/subagent-level agents were paused on 2026-06-26. Dept heads are idle, not paused. The pre-check that reads .status from the agent list will correctly block dispatch to any paused agent.
Codex adapter concurrency. Each Paperclip agent runs a Codex local adapter. If multiple agents are dispatched simultaneously, they compete for the same Codex subprocess slots on the VDS. The bridge should serialize or pace dispatches to avoid thrashing. For v1, dispatch one agent at a time and wait for idle confirmation before dispatching a second.
No native output callback. There is no webhook or callback from Paperclip to Hermes when an agent completes its task. Donna must poll paperclipai agent list for status change or use paperclipai issue list to read the completed issue. For v1, a manual "check results" step is acceptable. A polling skill (check-paperclip-result) is a follow-up item.
| File | Action | Founder gate |
|---|---|---|
~/.hermes/profiles/aina-chief-of-staff/skills/dispatch-to-paperclip/SKILL.md |
Create | No — Donna reads it, but the skill itself is inert until bridge.env is enabled |
/home/ali/.paperclip/instances/default/bridge.env |
Create | Yes — provisioning keys and setting enabled=true are founder-gated |
~/.hermes/profiles/paperclip-dispatcher/config.yaml |
Create (optional, phase 2) | No for creation; yes for activating via kanban default_assignee |
No changes to existing files are required for v1 direct dispatch. The kanban config.yaml in the Hermes profile does NOT need to change for the direct-dispatch path.
@"/Users/Ali/Downloads/PostHog and Mission Control (1).md" Check this and let's then check the work that was done by Finch, and the agent roster and SDLC plan and I had asked hermes to reconcile all linear issues that are pending so that we can map it to and assign it to the agents - will need a triage and roadmap agent to ensure that the loop works and agents are working on things that are not gated by me - for now let's map those and ensure all of them have the gpt codex config (which I believe they currently do to keep claude auth right now with us to save tokens and limits
The bridge spec came back complete and excellent (I've got it) — and it surfaced the real activation prerequisite: no dept-head has an API key issued yet, so provisioning keys is the true founder gate, not just a flag. I paused the build per your pivot and preserved the spec. Now to your actual direction — which is exactly the right next thing: the loop has no work source, and you want the real Linear backlog mapped onto the agents so they can run on things you're not gating.
{
"title": "Linear→agent mapping + telemetry plane",
"summary": "Map pending Linear issues to the agent roster (the loop's work source), verify all agents on codex/gpt config, scope a triage+roadmap agent, and absorb the PostHog-central / Mission-Control-cockpit architecture"
}Recording the bridge spec + roster durably and launching the investigation for your "for now" tasks:
cd /Users/Ali/PKM-monorepo cat >> docs/pkm-deepdive/RUNLOG.md <<'EOF' ## 2026-06-29 — Bridge spec returned + Ali pivot: Linear→agent mapping + telemetry plane - **Dispatch-bridge BUILD SPEC returned** (lane `ad30562`; full text in task output `ad30562cdf9375489.output`). Dispatch verb: `paperclipai agent-prompt --api-base http://127.0.0.1:3100 [--no-wake] [--json] <urlKey> <agentApiKey> <prompt>`. Files: new Hermes skill `dispatch-to-paperclip` + `/home/ali/.paperclip/instances/ default/bridge.env` (activation switch `PAPERCLIP_DISPATCH_ENABLED=false` default + `PAPERCLIP_DISPATCH_DRY_RUN`). **KEY BLOCKER: no dept-head has an API key issued yet** (`paperclipai token agent create`) → provisioning keys is the REAL activation gate, not just the flag. Dry-run test defined. **Build PAUSED per Ali pivot** (authorized, resume later). - **Roster (12 dept heads under Jessica=CEO/`jessica`, NOT a dispatch target):** Laurie(`laurie`,Data&Personalization), Monica(`monica`,Product&Learning), Mike(`mike`,Research&Intel), CCO(`chief-compliance-officer`,Compliance—PAUSED), Frodo(`frodo`,Release), Erlich(`erlich`,Media), Harvey(`harvey`,Growth), Benjamin(`benjamin`,Security&Privacy), Jared(`jared`,AgentOps), Richard(`richard`,Platform/CTO), Gandalf(`gandalf`,CoS-peer), Moneypenny(`moneypenny`,Concierge). - **Ali pivot (new direction):** (a) absorb PostHog/Mission-Control architecture doc (`~/Downloads/PostHog and Mission Control (1).md`): PostHog=central telemetry/data plane (6 mission events + OTLP logs + errors + AI-obs), Mission Control=thin Cloudflare PostHog-backed cockpit, Paperclip=orchestration SoT, Linear=human approval, Fusion=executor-only; (b) check Finch work + agent roster + SDLC plan; (c) find Hermes's Linear reconciliation (Ali asked Hermes to reconcile pending Linear issues); (d) **MAP pending Linear issues → agents** (the loop's missing work-source/ignition); (e) need a **triage + roadmap agent** to keep the loop fed with non-founder-gated work; (f) verify ALL agents on gpt/codex config (preserve Claude auth/tokens). - **Launched investigation workflow** (3 lanes: Finch+roster+SDLC / Hermes-Linear-reconcile+open-issues / codex-config-verify). - NEXT: synthesize → Linear→agent map + codex-config status + triage/roadmap-agent scope + PostHog/MC slotting. EOF git add docs/pkm-deepdive/RUNLOG.md git commit -q -m "docs(pkm-deepdive): RUNLOG — bridge spec returned + Ali pivot (Linear->agent map + telemetry plane) Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>" git push -q origin main && echo PUSHED
{
"script": "export const meta = {\n name: 'linear-to-agent-mapping',\n description: 'Map pending Linear issues to the agent roster: Finch work + roster + SDLC, Hermes Linear reconcile + open issues, codex-config verify',\n phases: [{ title: 'Investigate', detail: '3 read-only Sonnet lanes (finch+roster+sdlc / linear / codex-config)' }],\n}\n\nconst VDS = '/Users/Ali/PKM/scratch/pkm-deepdive/vds-exec.sh'\nconst CO = '7d58fc13-7c9a-4c87-b9f8-a7cfae8564a9'\n\nconst COMMON = `Read-only investigation lane. Goal: give Ali's agent org a WORK SOURCE — map the real pending Linear backlog onto the Paperclip dept-head agents so they run on things Ali isn't gating, and confirm they're all on the cost-efficient gpt/codex config (preserve Claude auth/tokens). Agents are PAUSED — do NOT wake/dispatch/change anything. VDS reads ONLY via ${VDS} run \"<cmd>\". Linear/GitHub via their MCP tools (load with ToolSearch). Company UUID = ${CO}. Cite exact paths/output. Return ONLY the structured object.`\n\nconst ROSTER_SCHEMA = {\n type: 'object', additionalProperties: false,\n required: ['finchWork','roster','orgStructure','sdlcPlan','canonForAgents'],\n properties: {\n finchWork: { type: 'string', description: 'What Finch built (/srv/aina/ops/finch-pkm-mapping/): the per-agent canon classification, REFERENCE_MAP, how it maps binding sources to each agent. Summarize + state freshness.' },\n roster: { type: 'array', items: { type: 'object', additionalProperties: true }, description: 'Every dept-head agent: {agent, urlKey, role, team, domain, status}. Get live via paperclipai agent list --company-id <CO> --json (through the wrapper). Include sub-agents if present.' },\n orgStructure: { type: 'string', description: 'The hierarchy: CEO (Jessica) -> dept heads -> reports. Who reports to whom.' },\n sdlcPlan: { type: 'string', description: 'Find the AINA SDLC plan (search /srv/repos, ~/Downloads/agentops for ali/, Linear docs, repo docs). Where it lives + the stages + how work flows idea->PRD->issue->agent->PR->release.' },\n canonForAgents: { type: 'string', description: 'How Finch maps canon/reference to each agent (the manifests) — and whether attached or staged.' },\n },\n}\n\nconst LINEAR_SCHEMA = {\n type: 'object', additionalProperties: false,\n required: ['hermesReconciliation','openIssues','countsByProject','founderGatedVsAgentDoable'],\n properties: {\n hermesReconciliation: { type: 'string', description: 'Ali asked Hermes to reconcile all pending Linear issues. FIND that work: Hermes state.db/sessions (ls ~/.hermes/; query state.db for recent linear/reconcile sessions via the wrapper), any report doc, or Linear comments. What did Hermes reconcile + the outcome + where it lives.' },\n openIssues: { type: 'array', items: { type: 'object', additionalProperties: true }, description: 'The PENDING/open Linear issues via the Linear MCP (ToolSearch \"linear list issues\"). Each: {id, identifier, title, state, project, team, labels, assignee}. Cover AIOPS/Agent-Memory, ACAD, DAT, PLAT, AIN, VISION. Cap at the most relevant ~60 if huge, and say so.' },\n countsByProject: { type: 'object', additionalProperties: true, description: 'open-issue counts per project/team' },\n founderGatedVsAgentDoable: { type: 'string', description: 'Which open issues are founder-gated (payments/secrets/legal/public-launch/approval) vs agent-doable now. Give the rough split + examples of each.' },\n },\n}\n\nconst CODEX_SCHEMA = {\n type: 'object', additionalProperties: false,\n required: ['perScope','allOnCodex','exceptions','recommendation'],\n properties: {\n perScope: { type: 'array', items: { type: 'object', additionalProperties: true }, description: 'Per scope {scope (paperclip-company/per-agent/hermes), engine, model, provider, auth, onCodex(bool), evidence(path)}. Read the Paperclip company codex-home config.toml + auth.json (~/.paperclip/instances/default/companies/<CO>/codex-home/), any per-agent overrides, and ~/.hermes/config.yaml (model/provider + aux slots).' },\n allOnCodex: { type: 'boolean', description: 'Are ALL agents + Hermes on openai-codex/gpt-5.5 (NOT Claude/Anthropic)?' },\n exceptions: { type: 'array', items: { type: 'string' }, description: 'Any agent/slot on Claude/Anthropic or misconfigured (would burn Claude tokens). Empty if all good.' },\n recommendation: { type: 'string', description: 'If exceptions exist, the exact config fix. If none, confirm cost-safe.' },\n },\n}\n\nphase('Investigate')\n\nconst [roster, linear, codex] = await parallel([\n () => agent(`${COMMON}\n\nLANE A — Finch's work + the agent roster + the SDLC plan.\n(1) Read Finch's outputs at /srv/aina/ops/finch-pkm-mapping/ (CLASSIFIED-INDEX.yml head, MASTER/REFERENCE_MAP, a sample agent CONTEXT-MANIFEST, REVIEW-PACKET) → summarize what Finch built + the canon-per-agent mapping + attached-vs-staged.\n(2) Get the LIVE roster: ${VDS} run \"paperclipai agent list --company-id ${CO} --json\" (and any team/structure list). Return agent→urlKey→role→team→domain→status for all dept heads (+ sub-agents).\n(3) Find + summarize the AINA SDLC plan (idea→PRD→Linear issue→agent goal→implementation→CI/evals→PR/proof→release→retro). Search /srv/repos, ~/Downloads/agentops for ali/, repo docs.`,\n { label: 'finch-roster-sdlc', schema: ROSTER_SCHEMA }),\n\n () => agent(`${COMMON}\n\nLANE B — Hermes's Linear reconciliation + the pending Linear backlog (the loop's missing work-source).\n(1) Ali asked Hermes to reconcile all pending Linear issues so they can be mapped/assigned to agents. FIND that reconciliation: ${VDS} run \"ls -la ~/.hermes/; sqlite3 ~/.hermes/state.db '.tables'\" then look for recent sessions/messages mentioning linear/reconcile/triage (query state.db). Also check for any reconciliation report doc on the VDS or in Linear comments. Summarize what Hermes did + the result + where it lives.\n(2) Enumerate the OPEN/pending Linear issues using the Linear MCP (ToolSearch \"linear list issues\" / \"select:...list_issues,list_projects,list_teams\"). Cover the AINA projects (AIOPS/Agent Memory, ACAD, DAT, PLAT, AIN, VISION). Return id/identifier/title/state/project/team/labels/assignee for each. Note founder-gated vs agent-doable.`,\n { label: 'linear-reconcile', schema: LINEAR_SCHEMA }),\n\n () => agent(`${COMMON}\n\nLANE C — Verify ALL agents are on the gpt/codex config (cost protection — keep Claude auth/tokens for Ali+lead).\nConfirm every Paperclip agent + Hermes runs on openai-codex / gpt-5.5, NOT Claude/Anthropic. Read (via the wrapper, read-only): the Paperclip company codex-home config.toml + auth.json at ~/.paperclip/instances/default/companies/${CO}/codex-home/ (model/provider/auth); any per-agent config overrides; and ~/.hermes/config.yaml (primary model/provider + the 13 aux slots). Report per scope with evidence, the allOnCodex boolean, and any exception that would burn Claude tokens. Do NOT change anything.`,\n { label: 'codex-config', schema: CODEX_SCHEMA }),\n])\n\nreturn { roster, linear, codex }\n"
}Workflow launched in background. Task ID: wwd3qzfgc
Summary: Map pending Linear issues to the agent roster: Finch work + roster + SDLC, Hermes Linear reconcile + open issues, codex-config verify
Transcript dir: /Users/Ali/.claude/projects/-Users-Ali-PKM/ea21dc9e-a684-460b-b5d8-244fa739715e/subagents/workflows/wf_b211038e-6d9
Script file: /Users/Ali/.claude/projects/-Users-Ali-PKM-monorepo/ea21dc9e-a684-460b-b5d8-244fa739715e/workflows/scripts/linear-to-agent-mapping-wf_b211038e-6d9.js
(Edit this file with Write/Edit and re-invoke Workflow with {scriptPath: "/Users/Ali/.claude/projects/-Users-Ali-PKM-monorepo/ea21dc9e-a684-460b-b5d8-244fa739715e/workflows/scripts/linear-to-agent-mapping-wf_b211038e-6d9.js"} to iterate without resending the script.)
Run ID: wf_b211038e-6d9
To resume after editing the script: Workflow({scriptPath: "/Users/Ali/.claude/projects/-Users-Ali-PKM-monorepo/ea21dc9e-a684-460b-b5d8-244fa739715e/workflows/scripts/linear-to-agent-mapping-wf_b211038e-6d9.js", resumeFromRunId: "wf_b211038e-6d9"}) — completed agents return cached results.
You will be notified when it completes. Use /workflows to watch live progress.While that investigation runs, my read on the PostHog + Mission Control doc — you asked me to check it, and it's a genuinely good call that fits exactly what we found today.
That doc (it's a reference/strategy output, not canon — I'm treating it as advisory) lands on: PostHog = central telemetry plane, Mission Control = thin cockpit on top of it, Paperclip = the work source-of-truth, Linear = human approvals, Fusion = executor-only. That's not just sound — it directly solves the visibility gap we found an hour ago. Remember the loop investigation: Mission Control today shows your paused roster, not the live loop; agent outputs aren't captured; the readiness tab is a frozen snapshot. The doc's answer — emit canonical mission events to PostHog, have Mission Control read them — is precisely how you get "see progress without reading code," with less VDS maintenance (no custom collectors to babysit).
The clean insight: the doc's 6 events (mission_started → task_planned → run_executed → artifact_emitted → release_promoted → mission_resolved) are the create→assign→run→capture→verify chain from the diagram I showed you — made observable. Its evidence_sha "only surface material changes" rule is tailor-made for how you want the cockpit to feel (signal, not noise).
So it supersedes one piece of our plan: Stage 0's "surface the loop in Mission Control" should be done PostHog-backed, not with bespoke VDS scrapers. Better target.
mission_id/run_id (not a "system" distinct-id) for agent events.How it connects to what you just asked for: the Linear→agent mapping gives the loop its work source; the dispatch bridge runs it; and the moment a dispatch fires mission_started, the loop becomes visible in Mission Control via PostHog. Work-source + telemetry are the two halves of "make the loop real and seeable." I'll fold the 6-event schema in as the canonical ops contract when we build the bridge.
Running now (investigation wwd3qzfgc): Finch's work + the full agent roster + the SDLC plan · finding Hermes's Linear reconciliation + enumerating every open Linear issue · and confirming all agents are on the gpt/codex config (your cost-protection check). When it lands I'll bring you the actual Linear-issue → agent map, the codex-config status, and a scope for the triage + roadmap agent that keeps the loop fed with work you're not gating. That triage/roadmap role is the real answer to "agents working on things not blocked by me."
Right — the doc's core schema treats Fusion/RunFusion as the runner and only bolts Paperclip on later. But Paperclip is now the orchestrator and the dept-head agents are what actually run the work, so the telemetry has to make the Paperclip agent execution layer first-class, with Fusion demoted to just one optional executor. Let me write that as the canonical contract.
# AINA Ops Telemetry Schema — Paperclip-primary (canonical ops contract)
> The emit contract for the operating loop. Adapts the reference PostHog/Mission-Control doc
> (`~/Downloads/PostHog and Mission Control (1).md`), which was Fusion/RunFusion-centric, to the
> **current reality: Paperclip is the orchestration source-of-truth and its department-head agents are the
> primary executors; Fusion is one optional executor.** PostHog = central plane; Mission Control reads it.
> Aligns with the AINA OS Reference (Paperclip=work layer, Linear=human approval, GitHub=audit+deploy only
> from `ainative-release`). 2026-06-29.
## Orchestration model (was Fusion-first; now Paperclip-first)
```
orchestrator : "paperclip" # Paperclip owns task/mission state — the SoT
coordinator : "hermes" # CoS routes/sequences; not the task DB
executor : "paperclip-agent" | "hermes" | "fusion" | "codex" | "claude-code"
| "github-actions" | "cloudflare-worker"
```
Fusion appears ONLY when `executor="fusion"` (a task that needs heavy/long/lab execution). Default executor for
dept-head work = `paperclip-agent` (a Paperclip agent running its Codex adapter). **This is the key correction:
the schema's default runner is a Paperclip dept-head agent, not a Fusion mission.**
## Events
### A. The 6 canonical MISSION events (kept — the spine, now Paperclip-aware)
`mission_started` · `task_planned` · `run_executed` · `artifact_emitted` · `release_promoted` · `mission_resolved`
— a `mission` = a unit of work (typically one Linear issue) flowing through the loop.
### B. NEW Paperclip AGENT-LAYER events (the dept-head execution lifecycle — what the reports showed)
The reports we audited (Benjamin/Silva/Frodo/Jared verdicts, PASS/FAIL, findings) had no telemetry. These make
that layer observable:
| Event | When | Key props (beyond common) |
|---|---|---|
| `agent_dispatched` | Hermes→Paperclip bridge files a task to a dept head | `dispatch_mode` (dry\|live), `source` (kanban\|triage\|ali), `paperclip_issue_id` |
| `agent_woke` | heartbeat wakeup triggered | `trigger`, `reason` |
| `agent_verdict` | a dept head emits a review/verification verdict | `verdict` (pass\|fail\|blocked), `findings_count`, `target` (pr\|commit\|issue), `evidence_sha` |
| `agent_blocked` | agent can't proceed | `blocker` (founder_gate\|api_unreachable\|missing_canon\|missing_key), `detail` |
| `agent_returned_idle` | agent finished, status→idle (the poll signal Hermes watches) | `outcome` |
## Common properties (every ops event) — Paperclip identity is first-class
```json
{
"surface": "ops|product|build|runtime|agent|security",
"project": "AINA",
"team": "platform|media|data|academy|security|growth|research|...",
"service": "hermes|paperclip|runfusion|mission-control|app|cloudflare-worker",
"orchestrator": "paperclip",
"coordinator": "hermes",
"executor": "paperclip-agent|hermes|fusion|codex|claude-code|github-actions|cloudflare-worker",
"paperclip_company_id": "7d58fc13-7c9a-4c87-b9f8-a7cfae8564a9",
"paperclip_agent_id": "<uuid>",
"paperclip_agent_urlkey": "jared|benjamin|richard|laurie|...",
"paperclip_agent_role": "agentops|security|platform|data|media|growth|research|release|product|concierge",
"paperclip_issue_id": "<paperclip issue id>",
"paperclip_task_id": "PC-...",
"paperclip_run_id": "<run id>",
"mission_id": "msn_...",
"linear_issue_id": "AIOPS-216",
"fusion_mission_id": "<only when executor=fusion>",
"repo": "ainative/...", "branch": "work/...", "commit_sha": "abc123",
"release_id": "2026-06-29-001",
"evidence_sha": "sha256:...", // semantic material-change dedupe (Mission Control shows first occurrence)
"artifact_ref": "r2://... | github-artifact-url", // raw blob stays in R2/PKM; PostHog holds pointer+sha+summary
"redaction_level": "none|hash|pointer|redacted",
"env": "vds|staging|prod",
"version": "git-tag-or-semver", "ts": "ISO8601", "trace_id": "trace_..."
}
```
## Emit points mapped to the create→assign→run→capture→verify loop
| Loop join | Event(s) | Emitter |
|---|---|---|
| **create** | `mission_started` | triage/roadmap agent or Ali intent (from a Linear issue) |
| **assign** | `agent_dispatched` (+ `task_planned`) | the Hermes→Paperclip dispatch bridge |
| **run** | `agent_woke` → `run_executed` (`executor=paperclip-agent`, `runner=codex`) | Paperclip agent adapter |
| **capture** | `artifact_emitted` (report/patch → R2/github; pointer+sha to PostHog) | the agent / capture hook |
| **verify** | `agent_verdict` → `mission_resolved` | the dept head + the verifier chain |
| **blocked (any stage)** | `agent_blocked` → surfaced in Mission Control; if `blocker=founder_gate`, routed to the founder rail | any stage |
## Join keys (Mission Control joins the loop on)
`paperclip_task_id` · `paperclip_issue_id` · `linear_issue_id` · `mission_id` · `github_commit_sha` · `evidence_sha`
## Identity rule (cost + correctness)
- **System/agent events:** `distinct_id = mission_id || paperclip_task_id || paperclip_run_id` (NEVER a shared
`"system"` id) and set `$process_person_profile: false` for high-volume agent events — avoids junk person
profiles + per-distinct-ID limit blowups.
- **Product/runtime events:** the real learner/user id + PostHog identity.
## What this unlocks (ties to the loop + Ali's "see progress")
- Mission Control's **Mission Timeline** = one `paperclip_task_id`/`mission_id` joining dispatched→run→artifact→
verdict→resolved + the Linear issue + the commit. That IS the no-code progress window.
- The **Agent Ops** page = per-dept-head throughput, verdicts (pass/fail), blockers, cost — so Ali sees which
agents are working and on what, without reading code.
- `agent_blocked` with `blocker=founder_gate` auto-populates the "What Needs You" founder rail — the loop tells
Ali exactly what only he can unblock.
## Status / sequencing
Contract only (this doc). **Nothing emits these yet.** Wire incrementally, cheapest-first: `agent_dispatched` +
`run_executed` + `agent_verdict` + `mission_resolved` from the dispatch bridge & reconciler when built → Mission
Control reads PostHog for the timeline. Raw bodies keep going to PKM/R2; PostHog stores pointer+sha+summary.
File created successfully at: /Users/Ali/PKM-monorepo/docs/pkm-deepdive/TELEMETRY-SCHEMA.md (file state is current in your context — no need to Read it back)
cd /Users/Ali/PKM-monorepo git add docs/pkm-deepdive/TELEMETRY-SCHEMA.md git commit -q -m "docs(pkm-deepdive): Paperclip-primary ops telemetry schema (not Fusion-centric) Adapts the reference PostHog schema to current reality: Paperclip = orchestration SoT + dept-head agents = primary executors; Fusion demoted to one optional executor. Keeps the 6 mission events, adds Paperclip agent-layer events (agent_dispatched/woke/verdict/blocked/returned_idle) + Paperclip identity props (company/agent/urlkey/role/issue/task/run) + emit points mapped to the loop + the founder-gate->'What Needs You' rail. Contract only; nothing emits yet (wire incrementally). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>" git push -q origin main && echo PUSHED
Done — rewrote the telemetry contract to be Paperclip-first (TELEMETRY-SCHEMA.md, landed). The shift, in plain terms:
Flipped the model. The reference doc made Fusion/RunFusion the thing that runs work and bolted Paperclip on as an afterthought. Reality is the opposite now: Paperclip orchestrates and owns the work, the dept-head agents are the runners, and Fusion is just one optional executor you reach for only on heavy/lab jobs. So orchestrator: "paperclip", default executor: "paperclip-agent", and fusion_mission_id only appears when Fusion is actually used.
Added the layer that was missing. Those dept-head reports we looked at — Benjamin's PASS, Silva's bug fix, Frodo's merge-readiness — produced zero telemetry. So I added the Paperclip agent-execution events: agent_dispatched → agent_woke → agent_verdict (pass/fail + findings) → agent_blocked → agent_returned_idle. Now when a dept head reviews something, an agent_verdict event fires with the result — that's what makes "which agent did what, and did it pass" visible.
Every event now carries Paperclip identity — which agent (paperclip_agent_urlkey: jared/benjamin/…), which role, which Paperclip issue/task — so Mission Control can show a per-dept-head view and a per-task timeline (dispatched → run → artifact → verdict → resolved, joined to the Linear issue and the commit).
The payoff for you specifically: an agent_blocked event tagged blocker=founder_gate auto-fills your "What Needs You" rail — the loop literally tells you the one thing only you can unblock, instead of you having to go find it. And nothing's wired yet, so this stays a cheap, incremental add (start with dispatched / run / verdict / resolved from the bridge + reconciler).
Meanwhile the investigation (wwd3qzfgc) is still running — the Linear→agent map, the codex-config check, Finch's work + the SDLC plan. That's the bigger deliverable; I'll bring it the moment it lands.
stale with superseded_by pointers; older PE/v1.1 + April ICP decisions historical; Brigade/salvage = historical by default; split Platform-Eng skill inheritance per-role (frontend/QA/payments/Cloudflare/API/docs/auth/reviewer/CTO no longer share one runtime/auth/billing bundle); removed frontend design-system refs from non-frontend Platform roles. Verification: 0 risky-canon refs across 29,351 entries and across all 62 manifests; 0 Calibrator sample-row failures; bash -n attach-manifests.sh passes; ./attach-manifests.sh --apply exits 10 (no founder token) / exits 12 (degraded PKM health) BEFORE any mutation or live-roster access.\n\nKEY CAVEAT: pkm-agent brain-health = DEGRADED at gather time (no fresh VDS snapshot receipt; receipts/PKM-BRAIN-HEALTH.md), so attach is hard-blocked. One unreachable source: /srv/archives/AINA-Salvage-2026-06-07. Finch made ZERO writes under /home/ali/PKM (read-only honored) and ran NO live attach.","roster":[{"agent":"Jessica","urlKey":"jessica","role":"ceo","title":"CEO / Mission Control","team":"command","reportsTo":null,"status":"idle","model":"gpt-5.5","adapterType":"codex_local","effort":"high","note":"company root; 12 direct reports (the dept heads)"},{"agent":"Gandalf","urlKey":"gandalf","role":"general","title":"Chief of Staff (peer to founder)","team":"command","reportsTo":"Jessica","status":"paused","model":"gpt-5.5","effort":"high","directReports":1},{"agent":"Moneypenny","urlKey":"moneypenny","role":"general","title":"Concierge (run-mode)","team":"command","reportsTo":"Jessica","status":"paused","model":"gpt-5.5","effort":"medium","directReports":0},{"agent":"Frodo","urlKey":"frodo","role":"devops","title":"Release Manager","team":"command","reportsTo":"Jessica","status":"idle","model":"gpt-5.5","effort":"medium","directReports":0},{"agent":"Jared","urlKey":"jared","role":"pm","title":"AgentOps Director","team":"agentops-infra","reportsTo":"Jessica","status":"idle","model":"gpt-5.5","effort":"high","directReports":7,"subAgents":["Root/vds-workshop-foreman","Jarvis/runtime-operator","Q/workflow-designer","Rambo/incident-commander","The Machine/routine-operator","Argus/agentops-inspector","Skills Architect/skills-architect"]},{"agent":"Richard","urlKey":"richard","role":"cto","title":"CTO / Platform Architect","team":"platform-eng","reportsTo":"Jessica","status":"idle","model":"gpt-5.5","effort":"high","directReports":8,"subAgents":["api-backend-engineer","auth-identity-engineer","cloudflare-runtime-engineer","Gimli/code-reviewer","documentation-engineer","frontend-engineer","payments-entitlements-engineer","Éowyn/qa-automation-engineer"]},{"agent":"Monica","urlKey":"monica","role":"pm","title":"Chief Product & Learning Officer","team":"product-learning","reportsTo":"Jessica","status":"idle","model":"gpt-5.5","effort":"high","directReports":5,"subAgents":["curriculum-architect","assessment-rubric-designer","learner-experience-architect","Analyst/product-analytics-interpreter","mobile-app-lead-architect (→ has its own 2 sub-agents: product-designer-ux-researcher, react-native-expo-engineer)"]},{"agent":"Laurie","urlKey":"laurie","role":"pm","title":"Chief Data & Personalization Officer","team":"personalization-data","reportsTo":"Jessica","status":"idle","model":"gpt-5.5","effort":"high","directReports":5,"subAgents":["data-ingestion-engineer","dataset-provenance-steward","embeddings-retrieval-engineer","role-ontology-engineer","Calibrator/evaluation-judge"]},{"agent":"Erlich","urlKey":"erlich","role":"pm","title":"Chief Media Officer","team":"media-engine","reportsTo":"Jessica","status":"idle","model":"gpt-5.5","effort":"high","directReports":6,"subAgents":["media-engine-producer","render-pipeline-engineer","Explainer/scriptwriter-explainer","audio-tts-producer","Editorial Critic/editorial-critic","Amplifier/social-repurposer"]},{"agent":"Harvey","urlKey":"harvey","role":"pm","title":"Chief Growth Officer","team":"growth","reportsTo":"Jessica","status":"idle","model":"gpt-5.5","effort":"high","directReports":4,"subAgents":["seo-aio-visibility","social-content-strategist","email-substack-strategist","landing-offer-strategist"]},{"agent":"Benjamin","urlKey":"benjamin","role":"pm","title":"Chief Security & Privacy Officer","team":"security-privacy","reportsTo":"Jessica","status":"idle","model":"gpt-5.5","effort":"high","directReports":4,"subAgents":["Silva/security-reviewer (Warden)","secrets-access-steward","privacy-consent-steward","supply-chain-auditor"]},{"agent":"Chief Compliance Officer","urlKey":"chief-compliance-officer","role":"pm","title":"Chief Compliance Officer","team":"legal-compliance","reportsTo":"Jessica","status":"paused","model":"gpt-5.5","effort":"high","directReports":4,"subAgents":["data-protection-officer","eu-gdpr-compliance-specialist","us-ccpa-compliance-specialist","policy-terms-drafter"]},{"agent":"Mike","urlKey":"mike","role":"pm","title":"Research & Intelligence Director","team":"research-intel","reportsTo":"Jessica","status":"idle","model":"gpt-5.5","effort":"high","directReports":3,"subAgents":["Legolas/research-scout","Elrond/evidence-synthesizer","adtech-trend-radar"]},{"_meta":"COST CONFIG AUDIT — all 62 agents are model=gpt-5.5, adapterType=codex_local, dangerouslyBypassApprovalsAndSandbox=true. ZERO agents on Claude/Anthropic. metadata.modelRouting note (set by Ali 2026-06-25): 'All agents routed to GPT-5.5 via codex_local; leads use high effort, non-leads use medium effort.' Leads=high reasoning effort, non-leads=medium. STATUSES: 49 paused, 13 idle, 0 active. heartbeat.enabled=false on all (no autonomous wake). Claude auth/tokens fully preserved — confirmed cost-efficient.","_finchUrlKey":"finch (Knowledge & Canon Steward, reportsTo Gandalf, p
... (truncated 47879 chars, full result in /private/tmp/claude-502/-Users-Ali-PKM/ea21dc9e-a684-460b-b5d8-244fa739715e/tasks/wwd3qzfgc.output)
The investigation's back with the full org tree and — importantly — your cost-config answer is confirmed: all 62 agents are on gpt-5.5 via codex_local, zero on Claude/Anthropic, heartbeat off. Claude auth fully preserved. Let me pull the full result for the Linear backlog + Hermes's reconciliation.
{
"file_path": "/private/tmp/claude-502/-Users-Ali-PKM/ea21dc9e-a684-460b-b5d8-244fa739715e/tasks/wwd3qzfgc.output"
}<system-reminder>[Truncated: PARTIAL view — showing lines 1-1030 of 1504 total (31029 tokens, cap 25000). Call Read with offset=1031 limit=1030 for the next page, or Grep to find a specific section. Do NOT answer from this page alone if the answer may be further in the file.]</system-reminder>
1 {
2 "summary": "Map pending Linear issues to the agent roster: Finch work + roster + SDLC, Hermes Linear reconcile + open issues, codex-config verify",
3 "agentCount": 3,
4 "logs": [],
5 "result": {
6 "roster": {
7 "finchWork": "Finch's outputs live at /srv/aina/ops/finch-pkm-mapping/ on the VDS (Lane A, issue AIN-42 / AIOPS-223 \"AIN-47\"). Freshness: regenerated 2026-06-26 20:54 EDT; \"READY FOR REVIEW\" — staged for founder review, NOT self-certified, NOT applied. Finch is itself a Paperclip agent (urlKey=finch, \"Knowledge & Canon Steward\", reportsTo Gandalf, currently paused).\n\nWHAT FINCH BUILT (5-phase A→E, generator = scripts/generate_finch_outputs.py, deterministic regen):\n- MASTER-REFERENCE-INDEX.yml (10.5MB) + CLASSIFIED-INDEX.yml (14.7MB, 29,351 entries) — every reachable PKM/repo source gathered across 5 \"planes\": agentops(614), brigade(143), pkm-daily(19,123), pkm-wiki(5,819), repo-doc(3,652). Each entry classified canon / historical / not-canon / stale. Counts: canon=17, historical=4,738, not-canon=24,510, stale=86 (deliberately conservative — most things are NOT binding canon).\n- REFERENCE_MAP.yml (108KB) + REFERENCE_MAP.extended.yml (90KB) — a STAGED rebuild of the missing scoped-context-topology/REFERENCE_MAP.yml, organized by department, listing per-department shared_references with source_id + classification + disposition (almost always \"point-to-path\", i.e. pointer not inlined content).\n- agents/<slug>/CONTEXT-MANIFEST.yml — 62 per-agent manifests (one per live roster agent; matches agents.json expectation of 62).\n- teams/<dept>/REFERENCES.md — 11 dept reference files (agentops-infra, command, growth, legal-compliance, media-engine, personalization-data, platform-eng, product-learning, product-learning-mobile, research-intel, security-privacy).\n- Governance docs: DECISION-CONTRACTS.md, MUST-NEVER-RULES.md, FAILURE-MODES.md, REVIEW-PACKET-v2.md, CHANGES.md, VERIFICATION.md, GATHER-REPORT.md, PHASE-E-CLOSEOUT.md.\n- attach-manifests.sh — the (founder-gated, dry-run-default) live-attach tool.\n\nPHASE E incorporated two adversarial reviews (reviews/gimli-review.md = Gimli/Code-Reviewer P1 attach-safety findings; reviews/calibrator-review.md = Calibrator/Evaluation-Judge classification findings). Fixes: downgraded PKM operational files / rendered HTML companions / handoffs / plans / placeholders away from binding canon; marked superseded May Cloudflare/Python decisions `stale` with superseded_by pointers; older PE/v1.1 + April ICP decisions `historical`; Brigade/salvage = historical by default; split Platform-Eng skill inheritance per-role (frontend/QA/payments/Cloudflare/API/docs/auth/reviewer/CTO no longer share one runtime/auth/billing bundle); removed frontend design-system refs from non-frontend Platform roles. Verification: 0 risky-canon refs across 29,351 entries and across all 62 manifests; 0 Calibrator sample-row failures; bash -n attach-manifests.sh passes; ./attach-manifests.sh --apply exits 10 (no founder token) / exits 12 (degraded PKM health) BEFORE any mutation or live-roster access.\n\nKEY CAVEAT: pkm-agent brain-health = DEGRADED at gather time (no fresh VDS snapshot receipt; receipts/PKM-BRAIN-HEALTH.md), so attach is hard-blocked. One unreachable source: /srv/archives/AINA-Salvage-2026-06-07. Finch made ZERO writes under /home/ali/PKM (read-only honored) and ran NO live attach.",
8 "roster": [
9 {
10 "agent": "Jessica",
11 "urlKey": "jessica",
12 "role": "ceo",
13 "title": "CEO / Mission Control",
14 "team": "command",
15 "reportsTo": null,
16 "status": "idle",
17 "model": "gpt-5.5",
18 "adapterType": "codex_local",
19 "effort": "high",
20 "note": "company root; 12 direct reports (the dept heads)"
21 },
22 {
23 "agent": "Gandalf",
24 "urlKey": "gandalf",
25 "role": "general",
26 "title": "Chief of Staff (peer to founder)",
27 "team": "command",
28 "reportsTo": "Jessica",
29 "status": "paused",
30 "model": "gpt-5.5",
31 "effort": "high",
32 "directReports": 1
33 },
34 {
35 "agent": "Moneypenny",
36 "urlKey": "moneypenny",
37 "role": "general",
38 "title": "Concierge (run-mode)",
39 "team": "command",
40 "reportsTo": "Jessica",
41 "status": "paused",
42 "model": "gpt-5.5",
43 "effort": "medium",
44 "directReports": 0
45 },
46 {
47 "agent": "Frodo",
48 "urlKey": "frodo",
49 "role": "devops",
50 "title": "Release Manager",
51 "team": "command",
52 "reportsTo": "Jessica",
53 "status": "idle",
54 "model": "gpt-5.5",
55 "effort": "medium",
56 "directReports": 0
57 },
58 {
59 "agent": "Jared",
60 "urlKey": "jared",
61 "role": "pm",
62 "title": "AgentOps Director",
63 "team": "agentops-infra",
64 "reportsTo": "Jessica",
65 "status": "idle",
66 "model": "gpt-5.5",
67 "effort": "high",
68 "directReports": 7,
69 "subAgents": [
70 "Root/vds-workshop-foreman",
71 "Jarvis/runtime-operator",
72 "Q/workflow-designer",
73 "Rambo/incident-commander",
74 "The Machine/routine-operator",
75 "Argus/agentops-inspector",
76 "Skills Architect/skills-architect"
77 ]
78 },
79 {
80 "agent": "Richard",
81 "urlKey": "richard",
82 "role": "cto",
83 "title": "CTO / Platform Architect",
84 "team": "platform-eng",
85 "reportsTo": "Jessica",
86 "status": "idle",
87 "model": "gpt-5.5",
88 "effort": "high",
89 "directReports": 8,
90 "subAgents": [
91 "api-backend-engineer",
92 "auth-identity-engineer",
93 "cloudflare-runtime-engineer",
94 "Gimli/code-reviewer",
95 "documentation-engineer",
96 "frontend-engineer",
97 "payments-entitlements-engineer",
98 "Éowyn/qa-automation-engineer"
99 ]
100 },
101 {
102 "agent": "Monica",
103 "urlKey": "monica",
104 "role": "pm",
105 "title": "Chief Product & Learning Officer",
106 "team": "product-learning",
107 "reportsTo": "Jessica",
108 "status": "idle",
109 "model": "gpt-5.5",
110 "effort": "high",
111 "directReports": 5,
112 "subAgents": [
113 "curriculum-architect",
114 "assessment-rubric-designer",
115 "learner-experience-architect",
116 "Analyst/product-analytics-interpreter",
117 "mobile-app-lead-architect (→ has its own 2 sub-agents: product-designer-ux-researcher, react-native-expo-engineer)"
118 ]
119 },
120 {
121 "agent": "Laurie",
122 "urlKey": "laurie",
123 "role": "pm",
124 "title": "Chief Data & Personalization Officer",
125 "team": "personalization-data",
126 "reportsTo": "Jessica",
127 "status": "idle",
128 "model": "gpt-5.5",
129 "effort": "high",
130 "directReports": 5,
131 "subAgents": [
132 "data-ingestion-engineer",
133 "dataset-provenance-steward",
134 "embeddings-retrieval-engineer",
135 "role-ontology-engineer",
136 "Calibrator/evaluation-judge"
137 ]
138 },
139 {
140 "agent": "Erlich",
141 "urlKey": "erlich",
142 "role": "pm",
143 "title": "Chief Media Officer",
144 "team": "media-engine",
145 "reportsTo": "Jessica",
146 "status": "idle",
147 "model": "gpt-5.5",
148 "effort": "high",
149 "directReports": 6,
150 "subAgents": [
151 "media-engine-producer",
152 "render-pipeline-engineer",
153 "Explainer/scriptwriter-explainer",
154 "audio-tts-producer",
155 "Editorial Critic/editorial-critic",
156 "Amplifier/social-repurposer"
157 ]
158 },
159 {
160 "agent": "Harvey",
161 "urlKey": "harvey",
162 "role": "pm",
163 "title": "Chief Growth Officer",
164 "team": "growth",
165 "reportsTo": "Jessica",
166 "status": "idle",
167 "model": "gpt-5.5",
168 "effort": "high",
169 "directReports": 4,
170 "subAgents": [
171 "seo-aio-visibility",
172 "social-content-strategist",
173 "email-substack-strategist",
174 "landing-offer-strategist"
175 ]
176 },
177 {
178 "agent": "Benjamin",
179 "urlKey": "benjamin",
180 "role": "pm",
181 "title": "Chief Security & Privacy Officer",
182 "team": "security-privacy",
183 "reportsTo": "Jessica",
184 "status": "idle",
185 "model": "gpt-5.5",
186 "effort": "high",
187 "directReports": 4,
188 "subAgents": [
189 "Silva/security-reviewer (Warden)",
190 "secrets-access-steward",
191 "privacy-consent-steward",
192 "supply-chain-auditor"
193 ]
194 },
195 {
196 "agent": "Chief Compliance Officer",
197 "urlKey": "chief-compliance-officer",
198 "role": "pm",
199 "title": "Chief Compliance Officer",
200 "team": "legal-compliance",
201 "reportsTo": "Jessica",
202 "status": "paused",
203 "model": "gpt-5.5",
204 "effort": "high",
205 "directReports": 4,
206 "subAgents": [
207 "data-protection-officer",
208 "eu-gdpr-compliance-specialist",
209 "us-ccpa-compliance-specialist",
210 "policy-terms-drafter"
211 ]
212 },
213 {
214 "agent": "Mike",
215 "urlKey": "mike",
216 "role": "pm",
217 "title": "Research & Intelligence Director",
218 "team": "research-intel",
219 "reportsTo": "Jessica",
220 "status": "idle",
221 "model": "gpt-5.5",
222 "effort": "high",
223 "directReports": 3,
224 "subAgents": [
225 "Legolas/research-scout",
226 "Elrond/evidence-synthesizer",
227 "adtech-trend-radar"
228 ]
229 },
230 {
231 "_meta": "COST CONFIG AUDIT — all 62 agents are model=gpt-5.5, adapterType=codex_local, dangerouslyBypassApprovalsAndSandbox=true. ZERO agents on Claude/Anthropic. metadata.modelRouting note (set by Ali 2026-06-25): 'All agents routed to GPT-5.5 via codex_local; leads use high effort, non-leads use medium effort.' Leads=high reasoning effort, non-leads=medium. STATUSES: 49 paused, 13 idle, 0 active. heartbeat.enabled=false on all (no autonomous wake). Claude auth/tokens fully preserved — confirmed cost-efficient.",
232 "_finchUrlKey": "finch (Knowledge & Canon Steward, reportsTo Gandalf, paused) — the agent that produced the mapping"
233 }
234 ],
235 "orgStructure": "Three-tier hierarchy, 62 agents total, company UUID 7d58fc13-7c9a-4c87-b9f8-a7cfae8564a9.\n\nTIER 0 — CEO: Jessica (jessica, role=ceo, \"CEO / Mission Control\", reportsTo=null, idle). Org-chain health = healthy.\n\nTIER 1 — 12 direct reports to Jessica (the department heads):\n- Gandalf — Chief of Staff (command; 1 sub = Finch). paused\n- Moneypenny — Concierge / run-mode (command; 0 subs). paused\n- Frodo — Release Manager (command; 0 subs). idle\n- Jared — AgentOps Director (agentops-infra; 7 subs). idle\n- Richard — CTO / Platform Architect (platform-eng; 8 subs). idle\n- Monica — Chief Product & Learning Officer (product-learning; 5 subs incl. mobile lead). idle\n- Laurie — Chief Data & Personalization Officer (personalization-data; 5 subs). idle\n- Erlich — Chief Media Officer (media-engine; 6 subs). idle\n- Harvey — Chief Growth Officer (growth; 4 subs). idle\n- Benjamin — Chief Security & Privacy Officer (security-privacy; 4 subs). idle\n- Chief Compliance Officer (legal-compliance; 4 subs). paused\n- Mike — Research & Intelligence Director (research-intel; 3 subs). idle\n\nTIER 2 — individual-contributor / engineer / specialist agents reporting to their dept head (49 of them). One 4-deep chain exists: Jessica → Monica → Mobile App Lead/Architect → {Product Designer-UX Researcher, React Native/Expo Engineer} — i.e. the mobile sub-team nests one level deeper than the others.\n\nNote: several dept heads also carry an embedded adversarial/QA reviewer as a direct report (Gimli/code-reviewer under Richard, Calibrator/evaluation-judge under Laurie, Silva/security-reviewer under Benjamin, Editorial Critic under Erlich, Argus/inspector under Jared) — these are the in-org review/gate agents.\n\nAll chains report orgChainHealth=healthy with no invalid ancestors. Current run-state: all 62 PAUSED-or-idle, heartbeat disabled (no autonomous execution) — consistent with the brief's \"agents are PAUSED.\"",
236 "sdlcPlan": "Two layers exist and agree.\n\n(1) CANONICAL OPERATING LIFECYCLE — the AINA AgentOps plugin (local, authoritative for day-to-day routing): /Users/Ali/plugins/aina-agentops/. Entry skill: /Users/Ali/plugins/aina-agentops/skills/aina-agentops-core/SKILL.md. The 8-stage lifecycle is defined verbatim in /Users/Ali/plugins/aina-agentops/references/lifecycle.md:\n Stage 0 Daily start gate (recent sessions, open PRs, worktrees, main alignment) →\n 1 Idea capture →\n 2 Decision artifact (PRD / ADR / contract spec / data spec / runbook / no-build note) →\n 3 Linear commitment (initiative → project/slice → issue range → issues) →\n 4 GitHub range setup (range doc, nested AGENTS.md, contracts, schemas, validation notes) →\n 5 Agent execution (one desktop session per issue range; bounded subagents for explore/review) →\n 6 PR + CI (founder pack, preview proof, validation receipt, Codex/QA review, no open blockers) →\n 7 Staging proof (merge to main, staging deploy, proof artifact, gate status) →\n 8 Production + retro (manual prod approval, smoke proof, release notes, Linear update, learning capture).\nPlanning hierarchy: Initiative (founder) → Project/Slice (founder + lead agent) → Epic/Issue-Range (one desktop chat, 3-5 issues, one gate) → Issue (one task/agent, one branch, one PR, one reviewable behavior) → Release Gate (QA/founder real-env proof). Non-negotiables: vertical tracer-bullet slices; 1 issue = 1 branch = 1 PR = 1 builder + 1 reviewer; start from live repo/GitHub/Linear truth; builders never approve their own merge gate; \"no issue = no work\" unless Ali asks for exploration. The plugin's stage skills map 1:1 onto the brief's pipeline: $idea-to-prd, $prd-to-issues, $agent-goal-charter, $implementation-pr-flow, $ci-evals-release-gate, plus $session-closeout / retro and $pkm-evidence-recall + $repo-truth-evidence-table for evidence.\n\n(2) FUTURE-STATE SKILLS CATALOG (the gap-filling roadmap, on VDS): /srv/repos/agentops-for-ali/skills-and-plugins/sdlc-agent-skills-catalog.md (+ .html companion). It enumerates 12 phases (Ideation/framing → Planning/PRD → Goal-setting → Decomposition-to-Beads → Routing/dispatch → Issue creation & sync → Prompting → Implementation → Triage → Review → Handoff → CI/CD merge gates) with proposed skills/commands and P0/P1/P2 priorities, plugging into Goals / Beads / Fusion / Dreaming / Linear+GH. Thesis: AINA is strong in the middle of the lifecycle (per-domain explorer→builder→reviewer triads + orchestrator + PR reviewer); the thin parts are the EDGES — idea→committed goal, goal→Beads DAG, routing, and closing the loop through merge gates + learnings. P0 items: write-spec, author-goal, dr-gate (done-review gate), decompose-to-beads, route-from-beads, prompt-author, self-modification-guard, start-reconcile, standards-enforcer, merge-gate.\n\nHow work flows idea→release in practice: idea → decision artifact/PRD → Linear initiative/project/issue-range/issue (Linear is scope-of-record) → GitHub range setup with nested AGENTS.md/contracts (GitHub = implementation truth) → a Paperclip dept-head agent owns the Linear issue and dispatches scoped sub-agents → implementation on one-branch-one-PR → CI + Codex/QA adversarial review (Gimli/Calibrator/Silver-class reviewers are baked into the org) → preview/staging proof → founder-gated production + retro/learning-capture. The org's release agent (Frodo / release_marshal) owns the release-gate stage; AgentOps-Infra (Jared's team) owns routing/runtime/incident.",
237 "canonForAgents": "Finch maps canon→agent via per-agent CONTEXT-MANIFEST.yml files (one per agent at /srv/aina/ops/finch-pkm-mapping/agents/<slug>/CONTEXT-MANIFEST.yml) layered on top of per-department teams/<dept>/REFERENCES.md. Mechanism, per the sample (api_backend_engineer):\n\n- Every manifest carries the same AUTHORITY ORDER (the conflict-resolution chain, from DECISION-CONTRACTS.md): 1 Ali gate → 2 GitHub merged code → 3 Linear active state → 4 Paperclip runtime → 5 repo AGENTS/DOMAIN/REFERENCE → 6 PKM canonical wiki → 7 PKM raw sessions/transcripts → 8 scratch. So PKM is explicitly a *reference library BELOW* repo/founder/Linear truth — never execution truth.\n- pkm_references: a SCOPED list (most agents = 18 refs; mobile lead + RN/Expo = 11 due to tighter scope) of source_id + classification (canon/not-canon/stale/historical) + disposition. Disposition is almost always \"point-to-path\" — manifests POINT to canonical paths, they do NOT inline executable facts (thresholds/regexes/env flags) — enforced by a MUST-NEVER rule. Top canon pointers for nearly all agents: repo-doc:aina-academy:FOUNDER-DECISIONS.md, repo-doc:aina-academy:canon/CANON-INDEX.md, repo-doc:aina-data-engine-room:FOUNDER-DECISIONS.md (legal/security teams also get canon/DPA-CONSENT.md). Refs are scope-matched to the agent's department/role; role-specific deny filters strip cross-domain noise (e.g. non-frontend Platform roles no longer pull design-system/UI-kit refs).\n- Each manifest also lists: skills (role-specific, e.g. api_backend gets cloudflare-workers, d1-migration, code-reviewer, pr-writer...), hooks (decision-contract, handoff, verify-before-close — all status \"activation-phase\"), permissions (filesystem scoped + PKM read-only + live-agent-attachment forbidden + code_write allowed in repo scope + secrets gated), directories, scratch_policy.\n\nATTACHED vs STAGED — STAGED ONLY (not attached). Every one of the 62 manifests carries pkm_health: degraded, attach_allowed: false, and review_status: \"staged-only; not attached to live agent\", plus explicit attach_blockers (\"PKM health degraded; manifest staged for review only\" + \"Live UUID/paused-state readback required before attach\"). The attach-manifests.sh --apply path is founder-gated and fails closed: exit 10 without the AIN-47 founder token, exit 12 on degraded PKM health — both BEFORE any roster mutation. Finch's PHASE-E-CLOSEOUT records zero live attachments performed. IMPORTANT distinction: the live agents DO already carry their own AGENTS.md / DEPARTMENT.md / MEMORY.md instruction bundles (from the SEPARATE earlier company-package activation — see /srv/repos/agentops-for-ali/.../company-package/attach-bundles.log \"attach done: files_ok=224 fail=0\"). Finch's PKM-reference manifests are an ADDITIONAL, not-yet-merged context layer awaiting founder review + a fresh non-degraded PKM brain-health before attach."
238 },
239 "linear": {
240 "hermesReconciliation": "TWO related reconciliation passes exist; both live on the VDS (Hermes runs as \"Donna / AINA Chief of Staff\").\n\n(1) THE LINEAR FACTORY RECONCILIATION (the one Ali asked for — map all pending Linear issues so they can be routed to agents). Lives at `/srv/aina/ops/linear-factory-reconciliation/` on the VDS, run 2026-06-26 by Donna/Hermes CoS. Plan: `20260626T184005Z-plan.md`. Evidence root: `20260626T184855Z/` (preflight/, exports/linear/{raw,markdown,ali-provided}/, logs/). Scripts: `export_linear.py`, `reconcile_and_writeback.py`, `finalize_writeback_from_enriched.py`. WHAT IT DID: full Linear export to VDS → evidence-backed state reconciliation (cross-checked each issue vs GitHub PRs/branches, repo handoff files, Paperclip runs, Fusion task logs) → wrote a structured `<!-- AINA-CONTROL-RECONCILIATION -->` block into issue DESCRIPTIONS (verified state, evidence links, comment rollup, execution routing: candidate executor / verifier / safety gate, do-not-redo note). OUTCOME (from `logs/finalize_writeback.stdout`): 767 writebacks, 0 failures, 63 \"Donna radar\" matches; per-team all-issue counts AIOPS 225, DAT 97, ACAD 116, INFRA 48, BIZ 43, LEARNIN 43, PLAT 40, UIA 33, AIN 30, VID 22, MKTG 21, PKM 15, VISION 15, RADAR 9, GROWTH 6, AGFM 3, REV 1. Note: `reconcile_and_writeback.py` hit a `csv.DictWriter` AttributeError on the analysis-CSV step, but the finalize/writeback path completed (0 failures), so the description writebacks landed. The reconciliation is also surfaced in Linear as AIOPS-226 \"Review: Linear factory reconciliation + current Paperclip agent roster\" (In Progress, project Donna CoS) with a Cloudflare mobile review at aina-mobile-review.pages.dev/docs/reports/2026-06-26-linear-factory-reconciliation.html. The DAT-85 issue body shows a concrete example of the reconciliation block (state `verified_done_candidate`, evidence links, executor=Richard/verifier=Gimli-Frodo).\n\n(2) THE AGENT-LANE MAPPING / CHARTER (the follow-on: map the 62 Paperclip agents into routable lanes). Lives in repo `/srv/repos/aina-paperclip-agent-context/` (charter = `AGENTS.md`; mapping = `mappings/agent-context-map.{md,csv,json}` generated 2026-06-27; live fleet snapshot = `mappings/live-paperclip-fleet-snapshot.{csv,json}`). Captured in Hermes session `20260627_230139_974166` \"Agent Lanes Mapping Configuration\" (`~/.hermes/state.db`, 2026-06-28). RESULT: all 62 live Paperclip agents mapped into 9 lanes — growth-media 11, security-privacy-compliance 10, qa-release 8, product-platform 7, content-curriculum 6, executive-governance 6, agentops-tools 6, data-personalization 5, research-intelligence 3. Migration status: 56 `mapped-not-applied`, 6 `applied-live-verified` (the content-curriculum lane only: Curriculum Architect, Monica, Editorial Critic, Assessment & Rubric Designer, Learner Experience Architect, Explainer — moved to cwd `workspaces/content-curriculum` with compact AGENTS.md/DOMAIN.md, no-write smoke passed, receipt `mappings/content-curriculum-apply-receipt-20260627.md`). All 62 agents remain PAUSED, 0 live runs. One caveat flagged by Hermes: Finch still has `heartbeat.enabled: true` while paused (the lone heartbeat exception). So: the work-source mapping is BUILT (every pending issue carries an executor/verifier routing block; every agent has a lane + verifier + target cwd), but only 1 of 9 lanes is live-applied — the other 8 are mapped-not-applied awaiting application.",
241 "openIssues": [
242 {
243 "id": "AIOPS-206",
244 "identifier": "AIOPS-206",
245 "title": "Reconcile AINA skills, plugins, hooks, and gates for agent-company setup",
246 "state": "Todo",
247 "project": "Paperclip",
248 "team": "AgentOps",
249 "labels": [],
250 "assignee": "none",
251 "priority": "Urgent",
252 "disposition": "agent-doable"
253 },
254 {
255 "id": "AIOPS-204",
256 "identifier": "AIOPS-204",
257 "title": "Build Foreman + Agent Company plugin for Paperclip/RunFusion setup",
258 "state": "In Progress",
259 "project": "Paperclip",
260 "team": "AgentOps",
261 "labels": [],
262 "assignee": "none",
263 "priority": "Urgent",
264 "disposition": "agent-doable"
265 },
266 {
267 "id": "AIOPS-228",
268 "identifier": "AIOPS-228",
269 "title": "Enable Donna CoS v1 autonomy without making Ali the bottleneck",
270 "state": "In Progress",
271 "project": "Donna CoS",
272 "team": "AgentOps",
273 "labels": [],
274 "assignee": "none",
275 "priority": "Urgent",
276 "disposition": "agent-doable"
277 },
278 {
279 "id": "AIOPS-227",
280 "identifier": "AIOPS-227",
281 "title": "Donna + Finch: consume AINA OS docs and produce first execution plan",
282 "state": "In Progress",
283 "project": "Donna CoS",
284 "team": "AgentOps",
285 "labels": [],
286 "assignee": "none",
287 "priority": "High",
288 "disposition": "agent-doable"
289 },
290 {
291 "id": "AIOPS-226",
292 "identifier": "AIOPS-226",
293 "title": "Review: Linear factory reconciliation + current Paperclip agent roster",
294 "state": "In Progress",
295 "project": "Donna CoS",
296 "team": "AgentOps",
297 "labels": [],
298 "assignee": "none",
299 "priority": "Medium",
300 "disposition": "founder-gated (review)"
301 },
302 {
303 "id": "AIOPS-235",
304 "identifier": "AIOPS-235",
305 "title": "[AIOPS Factory] Harden Donna non-Slack VDS dashboard route",
306 "state": "In Review",
307 "project": "-",
308 "team": "AgentOps",
309 "labels": [
310 "parallel:safe",
311 "agent-ready",
312 "size:S",
313 "area:repo-devex",
314 "squad:ops"
315 ],
316 "assignee": "none",
317 "priority": "High",
318 "disposition": "agent-doable"
319 },
320 {
321 "id": "AIOPS-184",
322 "identifier": "AIOPS-184",
323 "title": "Agent learnings corpus audit mode for PKM transcripts",
324 "state": "In Review",
325 "project": "-",
326 "team": "AgentOps",
327 "labels": [],
328 "assignee": "none",
329 "priority": "High",
330 "disposition": "agent-doable"
331 },
332 {
333 "id": "AIOPS-186",
334 "identifier": "AIOPS-186",
335 "title": "Adopt local-first execution lane for AINA modules",
336 "state": "In Review",
337 "project": "-",
338 "team": "AgentOps",
339 "labels": [
340 "founder-review",
341 "type:decision"
342 ],
343 "assignee": "none",
344 "priority": "High",
345 "disposition": "founder-gated (decision)"
346 },
347 {
348 "id": "AIOPS-183",
349 "identifier": "AIOPS-183",
350 "title": "Build AINA Mission Control dashboard",
351 "state": "In Review",
352 "project": "-",
353 "team": "AgentOps",
354 "labels": [
355 "founder-review"
356 ],
357 "assignee": "none",
358 "priority": "High",
359 "disposition": "agent-doable (founder-review on close)"
360 },
361 {
362 "id": "AIOPS-182",
363 "identifier": "AIOPS-182",
364 "title": "[about-ali] Review Codex draft and synthesize standing agent manual",
365 "state": "In Review",
366 "project": "-",
367 "team": "AgentOps",
368 "labels": [
369 "squad:docs",
370 "parallel:human-gate",
371 "founder-review"
372 ],
373 "assignee": "none",
374 "priority": "High",
375 "disposition": "founder-gated (human-gate)"
376 },
377 {
378 "id": "AIOPS-181",
379 "identifier": "AIOPS-181",
380 "title": "Instruction-file governance: lock CLAUDE.md/AGENTS.md baseline + drift-control",
381 "state": "In Review",
382 "project": "-",
383 "team": "AgentOps",
384 "labels": [
385 "area:repo-devex",
386 "priority:p0",
387 "agent-ready",
388 "squad:ops"
389 ],
390 "assignee": "none",
391 "priority": "High",
392 "disposition": "agent-doable"
393 },
394 {
395 "id": "AIOPS-197",
396 "identifier": "AIOPS-197",
397 "title": "[VDS] Repair cloud-home parity after Contabo seed",
398 "state": "In Review",
399 "project": "-",
400 "team": "AgentOps",
401 "labels": [
402 "agent-ready",
403 "area:repo-devex"
404 ],
405 "assignee": "none",
406 "priority": "High",
407 "disposition": "agent-doable"
408 },
409 {
410 "id": "AIOPS-192",
411 "identifier": "AIOPS-192",
412 "title": "Branch cleanup completion — collapse all AINA repos to main + only parked #302",
413 "state": "In Review",
414 "project": "-",
415 "team": "AgentOps",
416 "labels": [
417 "area:repo-devex",
418 "agent-ready",
419 "squad:ops"
420 ],
421 "assignee": "none",
422 "priority": "High",
423 "disposition": "agent-doable"
424 },
425 {
426 "id": "AIOPS-35",
427 "identifier": "AIOPS-35",
428 "title": "[VDS Cutover] Observability credential and event smoke",
429 "state": "In Review",
430 "project": "-",
431 "team": "AgentOps",
432 "labels": [
433 "area:analytics",
434 "parallel:human-gate",
435 "agent-ready"
436 ],
437 "assignee": "none",
438 "priority": "High",
439 "disposition": "founder-gated (secrets/credentials)"
440 },
441 {
442 "id": "AIOPS-34",
443 "identifier": "AIOPS-34",
444 "title": "[VDS Cutover] Session collision smoke",
445 "state": "In Review",
446 "project": "-",
447 "team": "AgentOps",
448 "labels": [
449 "area:repo-devex",
450 "agent-ready",
451 "squad:ops"
452 ],
453 "assignee": "none",
454 "priority": "High",
455 "disposition": "agent-doable"
456 },
457 {
458 "id": "AIOPS-33",
459 "identifier": "AIOPS-33",
460 "title": "[VDS Cutover] Golden image and recovery runbook",
461 "state": "In Review",
462 "project": "-",
463 "team": "AgentOps",
464 "labels": [
465 "parallel:human-gate",
466 "priority:p0",
467 "agent-ready",
468 "squad:ops"
469 ],
470 "assignee": "none",
471 "priority": "High",
472 "disposition": "agent-doable"
473 },
474 {
475 "id": "AIOPS-18",
476 "identifier": "AIOPS-18",
477 "title": "[Hermes] Create minimal AINA skill/prompt pack for intake, routing, summaries",
478 "state": "In Review",
479 "project": "-",
480 "team": "AgentOps",
481 "labels": [
482 "agent-ready",
483 "squad:ops"
484 ],
485 "assignee": "none",
486 "priority": "High",
487 "disposition": "agent-doable"
488 },
489 {
490 "id": "AIOPS-22",
491 "identifier": "AIOPS-22",
492 "title": "[Hermes] Safe initialization on VDS with least-privilege config",
493 "state": "In Review",
494 "project": "-",
495 "team": "AgentOps",
496 "labels": [
497 "agent-ready",
498 "squad:ops"
499 ],
500 "assignee": "none",
501 "priority": "High",
502 "disposition": "agent-doable"
503 },
504 {
505 "id": "AIOPS-193",
506 "identifier": "AIOPS-193",
507 "title": "BETA-LIVE-09: Beta cohort invite, support, and rollback runbook",
508 "state": "In Review",
509 "project": "-",
510 "team": "AgentOps",
511 "labels": [
512 "agent-ready",
513 "squad:ops",
514 "founder-review"
515 ],
516 "assignee": "none",
517 "priority": "High",
518 "disposition": "founder-gated (beta launch)"
519 },
520 {
521 "id": "AIOPS-205",
522 "identifier": "AIOPS-205",
523 "title": "Reduce Paperclip company-run context/token load with scoped source packs",
524 "state": "Todo",
525 "project": "Paperclip",
526 "team": "AgentOps",
527 "labels": [],
528 "assignee": "Ali Mehdi Mukadam",
529 "priority": "-",
530 "disposition": "agent-doable"
531 },
532 {
533 "id": "AIOPS-203(approx)",
534 "identifier": "AIOPS-xxx",
535 "title": "Design Donna/Paperclip internal patrol plugin or routine",
536 "state": "Todo",
537 "project": "Paperclip",
538 "team": "AgentOps",
539 "labels": [],
540 "assignee": "none",
541 "priority": "-",
542 "disposition": "agent-doable"
543 },
544 {
545 "id": "AIOPS-247",
546 "identifier": "AIOPS-247",
547 "title": "Paperclip report: Jessica / Mission Control (AIN-52)",
548 "state": "Todo",
549 "project": "Paperclip",
550 "team": "AgentOps",
551 "labels": [],
552 "assignee": "none",
553 "priority": "-",
554 "disposition": "agent-doable (report ingest)"
555 },
556 {
557 "id": "AIOPS-248",
558 "identifier": "AIOPS-248",
559 "title": "Paperclip report: Richard — CTO / Platform Architect (AIN-53)",
560 "state": "Todo",
561 "project": "Paperclip",
562 "team": "AgentOps",
563 "labels": [],
564 "assignee": "none",
565 "priority": "-",
566 "disposition": "agent-doable (report ingest)"
567 },
568 {
569 "id": "AIOPS-249",
570 "identifier": "AIOPS-249",
571 "title": "Paperclip report: Jared — AgentOps Director (AIN-54)",
572 "state": "Todo",
573 "project": "Paperclip",
574 "team": "AgentOps",
575 "labels": [],
576 "assignee": "none",
577 "priority": "-",
578 "disposition": "agent-doable (report ingest)"
579 },
580 {
581 "id": "AIOPS-250",
582 "identifier": "AIOPS-250",
583 "title": "Paperclip report: Laurie — Data & Personalization (AIN-55)",
584 "state": "Todo",
585 "project": "Paperclip",
586 "team": "AgentOps",
587 "labels": [],
588 "assignee": "none",
589 "priority": "Medium",
590 "disposition": "agent-doable (report ingest)"
591 },
592 {
593 "id": "AIOPS-dept-reports",
594 "identifier": "AIOPS-251..263",
595 "title": "Paperclip dept-head reports: Mike(AIN-57), Benjamin(AIN-58), Harvey(AIN-59), Erlich(AIN-60), Root(AIN-61), Documentation Engineer(AIN-62), Gimli(AIN-63), Frodo(AIN-64), Monica(AIN-65), Gimli-Addendum(AIN-66), Frodo-Addendum(AIN-67), + dept index AIN-52",
596 "state": "Todo",
597 "project": "Paperclip",
598 "team": "AgentOps",
599 "labels": [],
600 "assignee": "none",
601 "priority": "-",
602 "disposition": "agent-doable (report ingest)"
603 },
604 {
605 "id": "AIOPS-AgentObs",
606 "identifier": "AIOPS-xxx",
607 "title": "AINA Agent Observability Readiness: Opik, Cloudflare Workers AI, AI Gateway",
608 "state": "Todo",
609 "project": "Paperclip",
610 "team": "AgentOps",
611 "labels": [],
612 "assignee": "none",
613 "priority": "-",
614 "disposition": "agent-doable"
615 },
616 {
617 "id": "AIOPS-hermes-ssh",
618 "identifier": "AIOPS-xxx",
619 "title": "Review Hermes gateway VDS connection via SSH topology",
620 "state": "Todo",
621 "project": "Paperclip",
622 "team": "AgentOps",
623 "labels": [],
624 "assignee": "Ali Mehdi Mukadam",
625 "priority": "-",
626 "disposition": "agent-doable"
627 },
628 {
629 "id": "AIOPS-paperclip-heartbeat",
630 "identifier": "AIOPS-xxx",
631 "title": "Investigate Paperclip heartbeat API reliability from agent runs",
632 "state": "Todo",
633 "project": "Paperclip",
634 "team": "AgentOps",
635 "labels": [],
636 "assignee": "none",
637 "priority": "-",
638 "disposition": "agent-doable"
639 },
640 {
641 "id": "AIOPS-assignee-rules",
642 "identifier": "AIOPS-xxx",
643 "title": "Encode Paperclip assignee-boundary closeout rules + Implement assignee mention/wake protocol",
644 "state": "Todo",
645 "project": "Paperclip",
646 "team": "AgentOps",
647 "labels": [],
648 "assignee": "none",
649 "priority": "-",
650 "disposition": "agent-doable"
651 },
652 {
653 "id": "AGENT-MEM-inventory",
654 "identifier": "AIOPS-xxx",
655 "title": "Inventory and cross-link PKM / legacy memory work into Agent Memory",
656 "state": "Todo",
657 "project": "Agent Memory",
658 "team": "AgentOps",
659 "labels": [],
660 "assignee": "Ali Mehdi Mukadam",
661 "priority": "-",
662 "disposition": "agent-doable"
663 },
664 {
665 "id": "AGENT-MEM-mine",
666 "identifier": "AIOPS-xxx",
667 "title": "Mine PKM sessions & transcripts to extract skills, AGENTS.md rules, and agent patterns",
668 "state": "Todo",
669 "project": "-",
670 "team": "AgentOps",
671 "labels": [
672 "agent-ready"
673 ],
674 "assignee": "Ali Mehdi Mukadam",
675 "priority": "-",
676 "disposition": "agent-doable"
677 },
678 {
679 "id": "AGENT-MEM-access",
680 "identifier": "AIOPS-xxx",
681 "title": "Design PKM-backed Agent Memory access model for Paperclip/RunFusion/Hermes",
682 "state": "Todo",
683 "project": "Agent Memory",
684 "team": "AgentOps",
685 "labels": [],
686 "assignee": "Ali Mehdi Mukadam",
687 "priority": "-",
688 "disposition": "agent-doable"
689 },
690 {
691 "id": "AIOPS-factory-drift",
692 "identifier": "AIOPS-xxx",
693 "title": "[AIOPS Factory] Reconcile Linear status drift and comment-only requirement",
694 "state": "Todo",
695 "project": "-",
696 "team": "AgentOps",
697 "labels": [
698 "parallel:safe",
699 "agent-ready",
700 "size:S",
701 "squad:ops"
702 ],
703 "assignee": "none",
704 "priority": "-",
705 "disposition": "agent-doable"
706 },
707 {
708 "id": "AIOPS-runfusion-import",
709 "identifier": "AIOPS-xxx",
710 "title": "Run AINA Data Import mission via Fusion — Stage A (supervised)",
711 "state": "Todo",
712 "project": "RunFusion",
713 "team": "AgentOps",
714 "labels": [],
715 "assignee": "none",
716 "priority": "-",
717 "disposition": "founder-gated (supervised Fusion run)"
718 },
719 {
720 "id": "AIOPS-vds-mcp",
721 "identifier": "AIOPS-xxx",
722 "title": "VDS lane: Dedicated MCP architecture local-first build",
723 "state": "Todo",
724 "project": "-",
725 "team": "AgentOps",
726 "labels": [],
727 "assignee": "none",
728 "priority": "-",
729 "disposition": "agent-doable"
730 },
731 {
732 "id": "AIOPS-vds-admin",
733 "identifier": "AIOPS-xxx",
734 "title": "VDS lane: Admin command center local-first build",
735 "state": "Todo",
736 "project": "-",
737 "team": "AgentOps",
738 "labels": [],
739 "assignee": "none",
740 "priority": "-",
741 "disposition": "agent-doable"
742 },
743 {
744 "id": "AIOPS-vds-analytics",
745 "identifier": "AIOPS-xxx",
746 "title": "VDS lane: Analytics and telemetry local-first build",
747 "state": "Todo",
748 "project": "-",
749 "team": "AgentOps",
750 "labels": [],
751 "assignee": "none",
752 "priority": "-",
753 "disposition": "agent-doable"
754 },
755 {
756 "id": "AIOPS-vds-inbox",
757 "identifier": "AIOPS-xxx",
758 "title": "VDS lane: Cloudflare agentic inbox experiments local-first build",
759 "state": "Todo",
760 "project": "-",
761 "team": "AgentOps",
762 "labels": [],
763 "assignee": "none",
764 "priority": "-",
765 "disposition": "agent-doable"
766 },
767 {
768 "id": "AIOPS-session-contract",
769 "identifier": "AIOPS-xxx",
770 "title": "P0: Session Contract + Diff Boundary Guard for Claude/Codex Desktop Worktrees",
771 "state": "Todo",
772 "project": "-",
773 "team": "AgentOps",
774 "labels": [
775 "area:repo-devex",
776 "priority:p0",
777 "agent:human",
778 "squad:ops"
779 ],
780 "assignee": "none",
781 "priority": "-",
782 "disposition": "founder-gated (agent:human)"
783 },
784 {
785 "id": "AIOPS-tessl",
786 "identifier": "AIOPS-xxx",
787 "title": "Tessl Skill Evaluator for AgentOps Skills & Plugins",
788 "state": "Todo",
789 "project": "-",
790 "team": "AgentOps",
791 "labels": [],
792 "assignee": "none",
793 "priority": "-",
794 "disposition": "agent-doable"
795 },
796 {
797 "id": "AIOPS-vds-pe",
798 "identifier": "AIOPS-xxx",
799 "title": "[AgentOps] VDS personalization-engine background processing lane",
800 "state": "Todo",
801 "project": "-",
802 "team": "AgentOps",
803 "labels": [
804 "squad:data",
805 "area:personalization",
806 "agent-ready"
807 ],
808 "assignee": "none",
809 "priority": "-",
810 "disposition": "agent-doable"
811 },
812 {
813 "id": "AIOPS-vds-video",
814 "identifier": "AIOPS-xxx",
815 "title": "[AgentOps] VDS video/media production lane using AINA lesson style system",
816 "state": "Todo",
817 "project": "-",
818 "team": "AgentOps",
819 "labels": [
820 "area:design",
821 "agent-ready",
822 "squad:ops"
823 ],
824 "assignee": "none",
825 "priority": "-",
826 "disposition": "agent-doable"
827 },
828 {
829 "id": "AIOPS-sp3b",
830 "identifier": "AIOPS-xxx",
831 "title": "SP3b — Fix Codex invocation (subprocess x64 Node blocker)",
832 "state": "Todo",
833 "project": "-",
834 "team": "AgentOps",
835 "labels": [
836 "agent-ready",
837 "agent:codex"
838 ],
839 "assignee": "none",
840 "priority": "-",
841 "disposition": "agent-doable"
842 },
843 {
844 "id": "ACAD-123",
845 "identifier": "ACAD-123",
846 "title": "Revise AINA curriculum for chat-first operator systems",
847 "state": "Todo",
848 "project": "-",
849 "team": "AINA Academy",
850 "labels": [],
851 "assignee": "none",
852 "priority": "High",
853 "disposition": "agent-doable (founder-decision-adjacent)"
854 },
855 {
856 "id": "ACAD-114",
857 "identifier": "ACAD-114",
858 "title": "Adaptive protection fires 0/20 — wire plateau/decelerate/fail-out/fallback",
859 "state": "Todo",
860 "project": "Production Lane B — Academy Product-Quality Loop",
861 "team": "AINA Academy",
862 "labels": [],
863 "assignee": "none",
864 "priority": "High",
865 "disposition": "agent-doable (trust-firewall, was the gate in CLAUDE.md guard)"
866 },
867 {
868 "id": "ACAD-112",
869 "identifier": "ACAD-112",
870 "title": "Calibration re-look: validate evaluator-leniency + L1-recovery-blend before cohort",
871 "state": "Todo",
872 "project": "Production Lane B — Academy Product-Quality Loop",
873 "team": "AINA Academy",
874 "labels": [],
875 "assignee": "none",
876 "priority": "Urgent",
877 "disposition": "agent-doable (trust-firewall, was the gate in CLAUDE.md guard)"
878 },
879 {
880 "id": "ACAD-69",
881 "identifier": "ACAD-69",
882 "title": "225 role-curriculum lessons are UNREVIEWED candidate drafts serving as production — QA/promote pass",
883 "state": "Todo",
884 "project": "Production Lane B — Academy Product-Quality Loop",
885 "team": "AINA Academy",
886 "labels": [],
887 "assignee": "none",
888 "priority": "High",
889 "disposition": "agent-doable (QA pass; founder promote gate at end)"
890 },
891 {
892 "id": "ACAD-60",
893 "identifier": "ACAD-60",
894 "title": "M3 · Routing-map ignition — populate learner capability_layer_status from matched role packet",
895 "state": "Todo",
896 "project": "Production Lane B — Academy Product-Quality Loop",
897 "team": "AINA Academy",
898 "labels": [],
899 "assignee": "none",
900 "priority": "High",
901 "disposition": "agent-doable"
902 },
903 {
904 "id": "ACAD-16",
905 "identifier": "ACAD-16",
906 "title": "VDS lane: Mobile app architecture local-first build",
907 "state": "Todo",
908 "project": "-",
909 "team": "AINA Academy",
910 "labels": [],
911 "assignee": "none",
912 "priority": "Medium",
913 "disposition": "agent-doable"
914 },
915 {
916 "id": "ACAD-13",
917 "identifier": "ACAD-13",
918 "title": "Define AINA Lab OS / Classroom Lab Mode for AI Fluent Pro",
919 "state": "Todo",
920 "project": "AINA Lab OS / Classroom Lab Mode Exploration",
921 "team": "AINA Academy",
922 "labels": [],
923 "assignee": "none",
924 "priority": "Medium",
925 "disposition": "founder-gated (pricing/cost analysis needed)"
926 },
927 {
928 "id": "ACAD-120",
929 "identifier": "ACAD-120",
930 "title": "Review AINA forensic audit and reconcile control-plane drift",
931 "state": "In Progress",
932 "project": "-",
933 "team": "AINA Academy",
934 "labels": [],
935 "assignee": "Ali Mehdi Mukadam",
936 "priority": "High",
937 "disposition": "founder-gated (Ali review)"
938 },
939 {
940 "id": "ACAD-67",
941 "identifier": "ACAD-67",
942 "title": "FOUNDER SESSION · Curriculum authoring + generator at scale",
943 "state": "In Progress",
944 "project": "Production Lane B — Academy Product-Quality Loop",
945 "team": "AINA Academy",
946 "labels": [],
947 "assignee": "none",
948 "priority": "Urgent",
949 "disposition": "founder-gated (founder session; largely decided)"
950 },
951 {
952 "id": "ACAD-1",
953 "identifier": "ACAD-1",
954 "title": "Convergence: aina-platform assets + engine-room data contract → aina-academy",
955 "state": "In Progress",
956 "project": "AINA Lab OS / Classroom Lab Mode Exploration",
957 "team": "AINA Academy",
958 "labels": [
959 "vds"
960 ],
961 "assignee": "none",
962 "priority": "High",
963 "disposition": "agent-doable"
964 },
965 {
966 "id": "ACAD-110",
967 "identifier": "ACAD-110",
968 "title": "Adapt DeepTutor patterns into AINA Learning OS",
969 "state": "In Progress",
970 "project": "-",
971 "team": "AINA Academy",
972 "labels": [],
973 "assignee": "none",
974 "priority": "Medium",
975 "disposition": "agent-doable"
976 },
977 {
978 "id": "ACAD-109",
979 "identifier": "ACAD-109",
980 "title": "AINA Live Readiness — CI + export-contract + DER bridge slice",
981 "state": "In Progress",
982 "project": "AINA Live Readiness — Governed Beta Bridge",
983 "team": "AINA Academy",
984 "labels": [],
985 "assignee": "none",
986 "priority": "High",
987 "disposition": "agent-doable (live-readiness gate at end is founder)"
988 },
989 {
990 "id": "ACAD-99",
991 "identifier": "ACAD-99",
992 "title": "Behavioral robustness follow-ups: mission-less lessons, templated coach feedback, grindable pass threshold",
993 "state": "In Review",
994 "project": "Production Lane B — Academy Product-Quality Loop",
995 "team": "AINA Academy",
996 "labels": [],
997 "assignee": "none",
998 "priority": "High",
999 "disposition": "agent-doable"
1000 },
1001 {
1002 "id": "ACAD-97",
1003 "identifier": "ACAD-97",
1004 "title": "Universal-base feed + enriched Universal track + Fluency Map product end",
1005 "state": "In Progress",
1006 "project": "Production Lane B — Academy Product-Quality Loop",
1007 "team": "AINA Academy",
1008 "labels": [],
1009 "assignee": "none",
1010 "priority": "High",
1011 "disposition": "agent-doable"
1012 },
1013 {
1014 "id": "ACAD-81",
1015 "identifier": "ACAD-81",
1016 "title": "[evaluator] Wire the 150 mastery-gate anchors into the grader request + dimension cross-walk",
1017 "state": "In Progress",
1018 "project": "Production Lane B — Academy Product-Quality Loop",
1019 "team": "AINA Academy",
1020 "labels": [],
1021 "assignee": "none",
1022 "priority": "High",
1023 "disposition": "agent-doable"
1024 },
1025 {
1026 "id": "PLAT-40",
1027 "identifier": "PLAT-40",
1028 "title": "Apply to Dodo Payments Startups",
1029 "state": "Todo",
1030 "project": "AINA Platform",This is a major unlock — the mapping you asked Hermes for is already done: a Linear Factory Reconciliation wrote evidence-backed routing blocks (executor/verifier/gate) into 767 issues with 0 failures, and all 62 agents are mapped into 9 lanes. You just couldn't see it. Let me read the rest (the full backlog + codex-config detail) before I synthesize.
{
"file_path": "/private/tmp/claude-502/-Users-Ali-PKM/ea21dc9e-a684-460b-b5d8-244fa739715e/tasks/wwd3qzfgc.output",
"offset": 1031,
"limit": 474
}1031 "team": "AINA Platform",
1032 "labels": [],
1033 "assignee": "Ali Mehdi Mukadam",
1034 "priority": "Medium",
1035 "disposition": "founder-gated (payments/account application)"
1036 },
1037 {
1038 "id": "PLAT-3",
1039 "identifier": "PLAT-3",
1040 "title": "fix: Front-end IA cleanup — unify nav, canonical CTA, retire /index-pre-v6 + waitlist hop (U7)",
1041 "state": "Todo",
1042 "project": "-",
1043 "team": "AINA Platform",
1044 "labels": [
1045 "agent:claude-code",
1046 "agent-ready",
1047 "parallel:safe"
1048 ],
1049 "assignee": "none",
1050 "priority": "Medium",
1051 "disposition": "agent-doable"
1052 },
1053 {
1054 "id": "PLAT-7",
1055 "identifier": "PLAT-7",
1056 "title": "BETA-LIVE-00: Limited preview launch tracer queue",
1057 "state": "In Progress",
1058 "project": "-",
1059 "team": "AINA Platform",
1060 "labels": [
1061 "type:slice",
1062 "priority:p0",
1063 "agent-ready",
1064 "squad:ops",
1065 "founder-review"
1066 ],
1067 "assignee": "none",
1068 "priority": "High",
1069 "disposition": "founder-gated (public/beta launch)"
1070 },
1071 {
1072 "id": "PLAT-6",
1073 "identifier": "PLAT-6",
1074 "title": "BETA-LIVE-01: Controlled public entry and limited-preview access gate",
1075 "state": "In Progress",
1076 "project": "-",
1077 "team": "AINA Platform",
1078 "labels": [
1079 "cloudflare-preview-required",
1080 "priority:p0",
1081 "status:blocked",
1082 "founder-review"
1083 ],
1084 "assignee": "none",
1085 "priority": "High",
1086 "disposition": "founder-gated (public launch; status:blocked)"
1087 },
1088 {
1089 "id": "PLAT-5",
1090 "identifier": "PLAT-5",
1091 "title": "BETA-LIVE-02: Anonymous assessment to score to email capture",
1092 "state": "In Progress",
1093 "project": "-",
1094 "team": "AINA Platform",
1095 "labels": [
1096 "area:assessment",
1097 "cloudflare-preview-required",
1098 "priority:p0",
1099 "status:blocked",
1100 "founder-review"
1101 ],
1102 "assignee": "none",
1103 "priority": "High",
1104 "disposition": "founder-gated (public launch; status:blocked)"
1105 },
1106 {
1107 "id": "PLAT-4",
1108 "identifier": "PLAT-4",
1109 "title": "BETA-LIVE-03: Score-to-Clerk signup bridge and user-row proof",
1110 "state": "In Progress",
1111 "project": "-",
1112 "team": "AINA Platform",
1113 "labels": [
1114 "area:auth",
1115 "cloudflare-preview-required",
1116 "priority:p0",
1117 "status:blocked",
1118 "founder-review"
1119 ],
1120 "assignee": "none",
1121 "priority": "High",
1122 "disposition": "founder-gated (public launch; status:blocked)"
1123 },
1124 {
1125 "id": "INFRA-43",
1126 "identifier": "INFRA-43",
1127 "title": "BETA-LIVE-08: Funnel analytics and operator evidence dashboard",
1128 "state": "In Review",
1129 "project": "-",
1130 "team": "Infrastructure",
1131 "labels": [
1132 "status:blocked"
1133 ],
1134 "assignee": "none",
1135 "priority": "High",
1136 "disposition": "agent-doable (blocked on beta-live chain)"
1137 },
1138 {
1139 "id": "INFRA-42",
1140 "identifier": "INFRA-42",
1141 "title": "BETA-LIVE-05: Runtime environment, D1 migrations, and binding proof",
1142 "state": "In Review",
1143 "project": "-",
1144 "team": "Infrastructure",
1145 "labels": [
1146 "cloudflare-preview-required",
1147 "priority:p0",
1148 "founder-review"
1149 ],
1150 "assignee": "none",
1151 "priority": "High",
1152 "disposition": "founder-gated (remote D1 migrations need approval)"
1153 },
1154 {
1155 "id": "LEARNIN-34",
1156 "identifier": "LEARNIN-34",
1157 "title": "BETA-LIVE-04: Two-lesson preview path with safe locked catalog",
1158 "state": "In Progress",
1159 "project": "-",
1160 "team": "Curriculum Learning Loop",
1161 "labels": [
1162 "csv-review-required",
1163 "cloudflare-preview-required",
1164 "priority:p0",
1165 "status:blocked",
1166 "founder-review"
1167 ],
1168 "assignee": "none",
1169 "priority": "High",
1170 "disposition": "founder-gated (public launch; status:blocked)"
1171 },
1172 {
1173 "id": "VISION-19",
1174 "identifier": "VISION-19",
1175 "title": "Doctrine revision: agnosticism (3 clauses) + open funding with evaluator firewall",
1176 "state": "Todo",
1177 "project": "-",
1178 "team": "AINA Vision Core",
1179 "labels": [],
1180 "assignee": "Ali Mehdi Mukadam",
1181 "priority": "Urgent",
1182 "disposition": "founder-gated (doctrine decision)"
1183 },
1184 {
1185 "id": "VISION-25",
1186 "identifier": "VISION-25",
1187 "title": "[DECISION] Pricing treatment of a future 51st spine capability",
1188 "state": "Todo",
1189 "project": "-",
1190 "team": "AINA Vision Core",
1191 "labels": [
1192 "parallel:human-gate",
1193 "decision-needed",
1194 "founder-review",
1195 "area:payments",
1196 "type:decision"
1197 ],
1198 "assignee": "none",
1199 "priority": "Medium",
1200 "disposition": "founder-gated (pricing decision)"
1201 },
1202 {
1203 "id": "VISION-23",
1204 "identifier": "VISION-23",
1205 "title": "Doctrine update: selector → gated generation (v3.1) + consolidation since v1.2",
1206 "state": "Todo",
1207 "project": "-",
1208 "team": "AINA Vision Core",
1209 "labels": [],
1210 "assignee": "none",
1211 "priority": "High",
1212 "disposition": "founder-gated (doctrine)"
1213 },
1214 {
1215 "id": "VISION-3",
1216 "identifier": "VISION-3",
1217 "title": "AINA North Star: Cloudflare-Native Learning OS Architecture",
1218 "state": "In Progress",
1219 "project": "AINA Lab OS / Classroom Lab Mode Exploration",
1220 "team": "AINA Vision Core",
1221 "labels": [],
1222 "assignee": "none",
1223 "priority": "No priority",
1224 "disposition": "founder-gated (vision/strategy doc)"
1225 },
1226 {
1227 "id": "VISION-6",
1228 "identifier": "VISION-6",
1229 "title": "AINA MCP Strategy: Distribution Before Destination",
1230 "state": "In Progress",
1231 "project": "AINA Lab OS / Classroom Lab Mode Exploration",
1232 "team": "AINA Vision Core",
1233 "labels": [],
1234 "assignee": "none",
1235 "priority": "No priority",
1236 "disposition": "founder-gated (strategy doc)"
1237 },
1238 {
1239 "id": "VISION-5",
1240 "identifier": "VISION-5",
1241 "title": "AINA Manifesto: AI Fluency to AI Native",
1242 "state": "In Progress",
1243 "project": "AINA Lab OS / Classroom Lab Mode Exploration",
1244 "team": "AINA Vision Core",
1245 "labels": [],
1246 "assignee": "none",
1247 "priority": "No priority",
1248 "disposition": "founder-gated (manifesto)"
1249 },
1250 {
1251 "id": "VISION-4",
1252 "identifier": "VISION-4",
1253 "title": "AINA Positioning: Driving School, Flight Simulator & AI Waste Reduction",
1254 "state": "In Progress",
1255 "project": "AINA Lab OS / Classroom Lab Mode Exploration",
1256 "team": "AINA Vision Core",
1257 "labels": [],
1258 "assignee": "none",
1259 "priority": "No priority",
1260 "disposition": "founder-gated (positioning)"
1261 },
1262 {
1263 "id": "VISION-24",
1264 "identifier": "VISION-24",
1265 "title": "AINA Constitution: Durable AI Fluency and Amplifying Intelligence",
1266 "state": "Backlog",
1267 "project": "-",
1268 "team": "AINA Vision Core",
1269 "labels": [],
1270 "assignee": "none",
1271 "priority": "High",
1272 "disposition": "founder-gated (constitution)"
1273 },
1274 {
1275 "id": "VISION-16",
1276 "identifier": "VISION-16",
1277 "title": "AINA Memory Architecture v1",
1278 "state": "Backlog",
1279 "project": "-",
1280 "team": "AINA Vision Core",
1281 "labels": [],
1282 "assignee": "none",
1283 "priority": "No priority",
1284 "disposition": "agent-doable (design; founder-ratify)"
1285 },
1286 {
1287 "id": "VISION-17",
1288 "identifier": "VISION-17",
1289 "title": "AINA Constitution v1 - Organizational Governance Layer",
1290 "state": "Backlog",
1291 "project": "-",
1292 "team": "AINA Vision Core",
1293 "labels": [],
1294 "assignee": "none",
1295 "priority": "No priority",
1296 "disposition": "founder-gated (governance)"
1297 },
1298 {
1299 "id": "VISION-18",
1300 "identifier": "VISION-18",
1301 "title": "AINA Soul.md v1 - Mission, Identity, Values & Founder Intent",
1302 "state": "Backlog",
1303 "project": "-",
1304 "team": "AINA Vision Core",
1305 "labels": [],
1306 "assignee": "none",
1307 "priority": "No priority",
1308 "disposition": "founder-gated (identity)"
1309 },
1310 {
1311 "id": "VISION-14",
1312 "identifier": "VISION-14",
1313 "title": "Positioning: The Driving School & Flight Simulator — competitive stance",
1314 "state": "Backlog",
1315 "project": "-",
1316 "team": "AINA Vision Core",
1317 "labels": [],
1318 "assignee": "none",
1319 "priority": "High",
1320 "disposition": "agent-doable (research; founder-ratify)"
1321 },
1322 {
1323 "id": "VISION-20",
1324 "identifier": "VISION-20",
1325 "title": "[DECISION 6] Vision-vs-beta — ratify the parking lot",
1326 "state": "Backlog",
1327 "project": "-",
1328 "team": "AINA Vision Core",
1329 "labels": [
1330 "parallel:human-gate",
1331 "decision-needed",
1332 "type:decision",
1333 "founder-review"
1334 ],
1335 "assignee": "none",
1336 "priority": "High",
1337 "disposition": "founder-gated (decision)"
1338 },
1339 {
1340 "id": "PKM-16",
1341 "identifier": "PKM-16",
1342 "title": "Add guarded R2 garbage collection for stale PKM body objects",
1343 "state": "Backlog",
1344 "project": "-",
1345 "team": "PKM",
1346 "labels": [],
1347 "assignee": "none",
1348 "priority": "Low",
1349 "disposition": "agent-doable (explicitly deferred 'do not implement now')"
1350 },
1351 {
1352 "id": "DAT-team",
1353 "identifier": "DAT-*",
1354 "title": "Data Engine (DAT) team — NO open issues. All backlog/unstarted/started queries returned empty; DAT-85/86/91/92 etc. all Done. DAT-94/96/97 referenced as blockers in DAT-85 comments but are themselves Done/closed per reconciliation.",
1355 "state": "all-Done",
1356 "project": "Production Lane C — Data Engine Supplier Readiness",
1357 "team": "Data Engine",
1358 "labels": [],
1359 "assignee": "none",
1360 "priority": "-",
1361 "disposition": "none-open"
1362 }
1363 ],
1364 "countsByProject": {
1365 "_method_note": "Counts are OPEN issues (Todo/In Progress/In Review/Backlog) enumerated via Linear MCP list_issues per team across unstarted+started(+backlog where relevant). Some AgentOps Paperclip-report rows share generic identifiers in my list because the single list_issues call truncated full ids; the dept-report set is ~13 issues (AIN-52..67 mirrored as AIOPS-247..263). The finalize_writeback.stdout per-team numbers (AIOPS 225, DAT 97, ACAD 116, etc.) are ALL-STATE totals incl. Done/Canceled, NOT open counts.",
1366 "AgentOps_AIOPS_open_total": "~44 (19 In Progress/In Review + ~25 Todo incl. ~13 Paperclip dept-head report ingests + Agent-Memory x3 + VDS-lane x6)",
1367 "AgentOps_by_project": {
1368 "Paperclip": "~30 (incl. 13 dept-report ingests, AIOPS-204/206 build+reconcile, observability, heartbeat, assignee-protocol)",
1369 "Donna_CoS": "3 (AIOPS-226/227/228)",
1370 "Agent_Memory": "3",
1371 "RunFusion": "1",
1372 "no-project_VDS+devex": "~7"
1373 },
1374 "AINA_Academy_ACAD_open": "15 (8 Todo/started in Production Lane B quality loop + ACAD-120 founder review + ACAD-1/110/109/97/81/99 + ACAD-16/13/123)",
1375 "AINA_Platform_PLAT_open": "6 (PLAT-3/40 Todo + PLAT-4/5/6/7 BETA-LIVE chain In Progress)",
1376 "Infrastructure_INFRA_open": "2 (INFRA-42/43, both BETA-LIVE chain In Review)",
1377 "Curriculum_Learning_Loop_LEARNIN_open": "1 shown (LEARNIN-34 BETA-LIVE-04); team has more all-state (43 total) — not fully enumerated this pass",
1378 "Data_Engine_DAT_open": "0 (fully closed; all open-state queries empty)",
1379 "AINA_Vision_Core_VISION_open": "13 (mostly founder-gated doctrine/strategy/constitution; VISION-19/21/22 decisions, 21/22 already Done)",
1380 "PKM_open": "1 (PKM-16 Backlog, explicitly deferred); PKM-2/3/4..14/15 all Done, PKM-1 Canceled",
1381 "BizOps_BIZ_open": "2 shown (BIZ-14, BIZ-37) — not fully enumerated (43 all-state)",
1382 "UI_and_UX_UIA_open": "1 shown (UIA-30 Playfair) — not fully enumerated (33 all-state)",
1383 "Marketing_MKTG_open": "1 shown (MKTG-14) — not fully enumerated (21 all-state)",
1384 "coverage_caveat": "Fully enumerated open backlog for the 6 requested projects (AIOPS/Agent-Memory, ACAD, DAT, PLAT, AIN-as-AgentOps/AINA-workspace, VISION). BIZ/UIA/MKTG/LEARNIN/GROWTH/VID/RADAR/REV/AGFM were only sampled (they surfaced via a cross-team 'AIN-' query); their full open counts were not enumerated since they fall outside the 6 named projects. The reconciliation's all-state team totals are listed above for scale."
1385 },
1386 "founderGatedVsAgentDoable": "ROUGH SPLIT of the enumerated open backlog (~80 issues across the 6 named projects): about 55-60% AGENT-DOABLE NOW, 40-45% FOUNDER-GATED. The work-source for the paused fleet is real and large — most of it does NOT need Ali.\n\nFOUNDER-GATED (do NOT route to agents without Ali) — the gates are payments, secrets, public/beta launch, doctrine/identity decisions, and named human-gate labels:\n- PUBLIC/BETA LAUNCH (the biggest gated cluster): PLAT-4/5/6/7, LEARNIN-34, INFRA-42 (remote D1 migrations), AIOPS-193 (beta cohort runbook) — all the BETA-LIVE-00..09 tracer chain, founder-review + status:blocked.\n- PAYMENTS / ACCOUNTS: PLAT-40 (Dodo Payments application), VISION-25 (pricing of 51st capability), ACAD-13 (Lab OS — needs model/evaluator/sandbox cost + pricing analysis).\n- SECRETS / CREDENTIALS: AIOPS-35 (observability credential smoke — auth/tokens).\n- DOCTRINE / IDENTITY / STRATEGY DECISIONS (Vision Core is almost entirely this): VISION-19/23/24/17/18/20/3/4/5/6, plus AIOPS-186 (local-first execution decision), AIOPS-182 (about-ali manual, parallel:human-gate), ACAD-67 (founder curriculum session), ACAD-120 + AIOPS-226 (forensic-audit/reconciliation reviews assigned to Ali).\n- EXPLICIT HUMAN-GATE LABEL: AIOPS-session-contract (agent:human), VISION-20/25 (parallel:human-gate), AIOPS-33 (parallel:human-gate but agent-ready body).\n- SUPERVISED RUN: the RunFusion Stage-A data-import mission (needs supervision/approval to run Fusion).\n\nAGENT-DOABLE NOW (safe to map onto the paused dept-head fleet — no Ali gate) — these are the work-source the loop is missing:\n- The two CLAUDE.md trust-firewall gate issues ACAD-112 (Urgent, calibration re-look) and ACAD-114 (adaptive-protection 0/20 wiring) are agent-doable engineering with proof attached — NOTE these were the redirect-gate in Ali's temporary CLAUDE.md guard; they remain open/unstarted, so per that guard new AINA product work should reconcile/route around them first.\n- Academy quality loop: ACAD-60/81/97/99/109/110/1 (engine wiring, evaluator anchors, universal feed, DER bridge), ACAD-69 (QA the 225 candidate lessons; only the final promote is founder).\n- AgentOps factory/infra: AIOPS-181 (instruction-file governance), AIOPS-192 (branch cleanup), AIOPS-197 (VDS parity), AIOPS-34 (session-collision smoke), AIOPS-33 (golden image), AIOPS-18/22 (Hermes skill pack + init), AIOPS-184 (learnings audit), AIOPS-235 (Donna dashboard route), AIOPS-204/206 (Foreman plugin + skills reconcile), AIOPS-205 (token-load reduction), the Paperclip heartbeat/assignee-protocol items, and the ~13 Paperclip dept-head report INGESTS (AIN-52..67 → mirror into Linear) which are pure agent formatting work.\n- Agent Memory x3 (inventory/mine/access-model) + the 4 VDS lanes (MCP, admin, analytics, agentic-inbox) + VDS personalization + VDS video lanes — all agent-ready, local-first.\n- PLAT-3 (front-end IA cleanup, agent:claude-code label), AIOPS-tessl, AIOPS-sp3b (Codex invocation fix).\n- Design-only/ratify-later: VISION-14/16 (research + memory-architecture design — agents can draft, Ali ratifies).\n\nROUTING MATCH (work-source → paused lane, per the agent-context-map): ACAD-* quality/engine work → product-platform lane (Richard, builders, cwd /srv/repos/aina-academy) + content-curriculum lane (Monica — the ONE live lane); AIOPS factory/devex → agentops-tools lane (Jared/Jarvis) + qa-release (Root/Gimli/Frodo); Agent-Memory/PKM-mining → data-personalization (Laurie) + research-intelligence (Mike); growth/media and security/compliance lanes stay draft/review-only until Ali opens public/legal lanes. CONFIG CONFIRMED COST-EFFICIENT: Hermes/Donna runtime default is gpt-5.5 on provider openai-codex (codex_gpt55_autoraise: true) and all 14 aux/sub-agent slots are openai-codex; Claude/Anthropic was removed from Hermes auth (auth.json.bak + config.yaml.bak both tagged 'remove-claude-anthropic', 2026-06-28), so the fleet runs on the GPT/Codex subscription and does NOT drain Claude tokens. All 62 Paperclip agents are paused, 0 live runs; only Finch still has heartbeat.enabled=true (flagged cleanup)."
1387 },
1388 "codex": {
1389 "perScope": [
1390 {
1391 "scope": "paperclip-company",
1392 "engine": "openai-codex (Codex CLI)",
1393 "model": "gpt-5.5",
1394 "provider": "openai (default — no model_provider key; Codex uses OpenAI provider)",
1395 "auth": "ChatGPT Pro OAuth (auth_mode=chatgpt, no OPENAI_API_KEY, no API billing)",
1396 "onCodex": true,
1397 "evidence": "VDS:~/.paperclip/instances/default/companies/7d58fc13-7c9a-4c87-b9f8-a7cfae8564a9/codex-home/config.toml line1 `model = \"gpt-5.5\"`; no `model_provider`/`[model_providers]` key; auth.json -> /home/ali/.codex/auth.json (auth_mode=chatgpt, chatgpt_plan_type=pro, iss=auth.openai.com, OPENAI_API_KEY present=False). Read via /Users/Ali/PKM/scratch/pkm-deepdive/vds-exec.sh run"
1398 },
1399 {
1400 "scope": "per-agent (Paperclip dept-head subagents)",
1401 "engine": "openai-codex (inherited)",
1402 "model": "gpt-5.5 (inherited from base)",
1403 "provider": "openai (inherited)",
1404 "auth": "same shared ChatGPT Pro OAuth",
1405 "onCodex": true,
1406 "evidence": "Same config.toml: NO `[agents.<name>]` named role tables and NO per-agent `*.config.toml` files exist (ls of *.config.toml = none). Inline comment lines 31-38 confirm: 'spawned subagents inherit the base config above (full-auto, danger-full-access, gpt-5.5/medium; plan mode high)'. [agents] block only sets max_threads=100/max_depth=4/job_max_runtime_seconds=7200. Anthropic scan: only benign hits are [projects.\"...Claude...\"] path entries, no Anthropic model/provider."
1407 },
1408 {
1409 "scope": "hermes (VDS — agent-org host)",
1410 "engine": "openai-codex",
1411 "model": "gpt-5.5 (primary) + all 13 aux slots model:'' (inherit gpt-5.5)",
1412 "provider": "openai-codex",
1413 "auth": "ChatGPT OAuth (active_provider=openai-codex, auth_mode=chatgpt); credential_pool=[copilot, openai-codex]; NO anthropic",
1414 "onCodex": true,
1415 "evidence": "VDS:~/.hermes/config.yaml lines1-3 model.default=gpt-5.5/provider=openai-codex; providers:{}; fallback_providers:[]; auxiliary block lines167-256 = 13 slots (vision,web_extract,compression,skills_hub,approval,mcp,title_generation,tts_audio_tags,triage_specifier,kanban_decomposer,profile_describer,curator,monitor) ALL provider:openai-codex,model:''; grep anthropic|claude|sonnet|opus|haiku = NONE outside comments. auth.json active_provider=openai-codex, providers=[openai-codex]. .env: no live (uncommented) Anthropic key, only Slack tokens live. Note: x_search uses grok-4.20-reasoning (xAI, not Claude). Backup files config.yaml.bak-*-remove-claude-anthropic show Claude was already removed 2026-06-28."
1416 },
1417 {
1418 "scope": "hermes (Mac — Ali+lead surface)",
1419 "engine": "openai-codex",
1420 "model": "gpt-5.5 (primary); 14 aux slots model:'' provider:auto",
1421 "provider": "openai-codex (primary); aux=auto -> resolves to openai-codex (no Anthropic cred to pick)",
1422 "auth": "ChatGPT OAuth (active_provider=openai-codex, auth_mode=chatgpt); credential_pool=[copilot, openai-codex]; NO anthropic cred",
1423 "onCodex": true,
1424 "evidence": "/Users/Ali/.hermes/config.yaml lines1-3 model.default=gpt-5.5/provider=openai-codex; providers:{}; fallback_providers:[]; auxiliary lines181-281 = 14 slots provider:auto/model:'' (adds background_review vs VDS). grep anthropic|claude = NONE outside comments. auth.json active_provider=openai-codex, providers=[openai-codex], credential_pool=[copilot,openai-codex] — NO anthropic credential exists, so 'auto' cannot route to Claude. .env: ALL LLM provider keys are commented out (no live ANTHROPIC_API_KEY); only SLACK_*_TOKEN live. config.yaml.bak-*-remove-claude-anthropic + auth.json.bak-*-remove-claude-anthropic confirm Claude was stripped."
1425 }
1426 ],
1427 "allOnCodex": true,
1428 "exceptions": [],
1429 "recommendation": "Cost-safe. Every scope runs on OpenAI/Codex (gpt-5.5) via ChatGPT Pro OAuth — zero Anthropic/Claude routing, so Paperclip agents + Hermes cannot burn Ali's Claude tokens. Specifics: (1) Paperclip company config.toml = gpt-5.5, no model_provider override (=OpenAI), no named [agents.*] tables and no per-agent *.config.toml, so all dept-head subagents inherit gpt-5.5; auth.json symlinks to shared ChatGPT-Pro OAuth (no OPENAI_API_KEY = subscription auth, not API billing). (2) VDS Hermes = gpt-5.5/openai-codex with all 13 aux slots pinned to openai-codex; fallback_providers empty; no Anthropic in config/auth/.env. (3) Mac Hermes primary = gpt-5.5/openai-codex; its 14 aux slots use provider:auto, BUT auto is harmless here because no Anthropic credential exists (auth.json providers=[openai-codex], credential_pool=[copilot,openai-codex], and all LLM keys in .env are commented out) — auto can only resolve to openai-codex/copilot. The *-remove-claude-anthropic config/auth backups on both hosts (2026-06-28) confirm a deliberate prior Claude strip. OPTIONAL HARDENING (not required, no token exposure today): to make the Mac match the VDS and remove any future risk if an Anthropic credential were ever added, change the 14 Mac Hermes aux slots from provider:auto to provider:openai-codex. Leave as-is otherwise — agents are PAUSED and nothing was changed."
1430 }
1431 },
1432 "workflowProgress": [
1433 {
1434 "type": "workflow_phase",
1435 "index": 1,
1436 "title": "Investigate"
1437 },
1438 {
1439 "type": "workflow_agent",
1440 "index": 1,
1441 "label": "finch-roster-sdlc",
1442 "phaseIndex": 1,
1443 "phaseTitle": "Investigate",
1444 "agentId": "aa608e88051b69aa3",
1445 "model": "claude-opus-4-8",
1446 "state": "done",
1447 "startedAt": 1782708359104,
1448 "queuedAt": 1782708359061,
1449 "attempt": 1,
1450 "lastToolName": "StructuredOutput",
1451 "lastToolSummary": "Finch's outputs live at /srv/aina/ops/finch-pkm-mapping/ on…",
1452 "promptPreview": "Read-only investigation lane. Goal: give Ali's agent org a WORK SOURCE — map the real pending Linear backlog onto the Paperclip dept-head agents so they run on things Ali isn't gating, and confirm they're all on the cost-efficient gpt/codex config (preserve Claude auth/tokens). Agents are PAUSED — do NOT wake/dispatch/change anything. VDS reads ONLY via /Users/Ali/PKM/scratch/pkm-deepdive/vds-exec…",
1453 "lastProgressAt": 1782708536092,
1454 "tokens": 96665,
1455 "toolCalls": 18,
1456 "durationMs": 176954,
1457 "resultPreview": "{\"finchWork\":\"Finch's outputs live at /srv/aina/ops/finch-pkm-mapping/ on the VDS (Lane A, issue AIN-42 / AIOPS-223 \\\"AIN-47\\\"). Freshness: regenerated 2026-06-26 20:54 EDT; \\\"READY FOR REVIEW\\\" — staged for founder review, NOT self-certified, NOT applied. Finch is itself a Paperclip agent (urlKey=finch, \\\"Knowledge & Canon Steward\\\", reportsTo Gandalf, currently paused).\\n\\nWHAT FINCH BUILT (5-ph…"
1458 },
1459 {
1460 "type": "workflow_agent",
1461 "index": 2,
1462 "label": "linear-reconcile",
1463 "phaseIndex": 1,
1464 "phaseTitle": "Investigate",
1465 "agentId": "a7b0be346a8f0a14c",
1466 "model": "claude-opus-4-8",
1467 "state": "done",
1468 "startedAt": 1782708359229,
1469 "queuedAt": 1782708359061,
1470 "attempt": 1,
1471 "lastToolName": "StructuredOutput",
1472 "lastToolSummary": "TWO related reconciliation passes exist; both live on the V…",
1473 "promptPreview": "Read-only investigation lane. Goal: give Ali's agent org a WORK SOURCE — map the real pending Linear backlog onto the Paperclip dept-head agents so they run on things Ali isn't gating, and confirm they're all on the cost-efficient gpt/codex config (preserve Claude auth/tokens). Agents are PAUSED — do NOT wake/dispatch/change anything. VDS reads ONLY via /Users/Ali/PKM/scratch/pkm-deepdive/vds-exec…",
1474 "lastProgressAt": 1782708704252,
1475 "tokens": 155417,
1476 "toolCalls": 38,
1477 "durationMs": 345023,
1478 "resultPreview": "{\"hermesReconciliation\":\"TWO related reconciliation passes exist; both live on the VDS (Hermes runs as \\\"Donna / AINA Chief of Staff\\\").\\n\\n(1) THE LINEAR FACTORY RECONCILIATION (the one Ali asked for — map all pending Linear issues so they can be routed to agents). Lives at `/srv/aina/ops/linear-factory-reconciliation/` on the VDS, run 2026-06-26 by Donna/Hermes CoS. Plan: `20260626T184005Z-plan.…"
1479 },
1480 {
1481 "type": "workflow_agent",
1482 "index": 3,
1483 "label": "codex-config",
1484 "phaseIndex": 1,
1485 "phaseTitle": "Investigate",
1486 "agentId": "a2fa7a452ae37383b",
1487 "model": "claude-opus-4-8",
1488 "state": "done",
1489 "startedAt": 1782708359270,
1490 "queuedAt": 1782708359061,
1491 "attempt": 1,
1492 "lastToolName": "StructuredOutput",
1493 "lastToolSummary": "Cost-safe. Every scope runs on OpenAI/Codex (gpt-5.5) via C…",
1494 "promptPreview": "Read-only investigation lane. Goal: give Ali's agent org a WORK SOURCE — map the real pending Linear backlog onto the Paperclip dept-head agents so they run on things Ali isn't gating, and confirm they're all on the cost-efficient gpt/codex config (preserve Claude auth/tokens). Agents are PAUSED — do NOT wake/dispatch/change anything. VDS reads ONLY via /Users/Ali/PKM/scratch/pkm-deepdive/vds-exec…",
1495 "lastProgressAt": 1782708543638,
1496 "tokens": 69080,
1497 "toolCalls": 21,
1498 "durationMs": 184368,
1499 "resultPreview": "{\"perScope\":[{\"scope\":\"paperclip-company\",\"engine\":\"openai-codex (Codex CLI)\",\"model\":\"gpt-5.5\",\"provider\":\"openai (default — no model_provider key; Codex uses OpenAI provider)\",\"auth\":\"ChatGPT Pro OAuth (auth_mode=chatgpt, no OPENAI_API_KEY, no API billing)\",\"onCodex\":true,\"evidence\":\"VDS:~/.paperclip/instances/default/companies/7d58fc13-7c9a-4c87-b9f8-a7cfae8564a9/codex-home/config.toml line1 `m…"
1500 }
1501 ],
1502 "totalTokens": 321162,
1503 "totalToolCalls": 77
1504 }Now I have the complete picture, and it's genuinely good news — most of what you asked for already exists; it's just staged and invisible. Let me land the synthesis durably, then show you.
# Agent Work Map — Linear backlog → agents, config, triage/roadmap (the loop's work-source) > Investigation `wwd3qzfgc` (Finch + roster + SDLC / Hermes Linear reconciliation + open backlog / codex-config). > Answers Ali's ask: map pending Linear issues to agents + confirm all agents are on the cost-efficient codex > config. Headline: **most of this is already DONE — it's staged + invisible, not missing.** 2026-06-29. ## TL;DR 1. **The mapping you asked Hermes for is DONE (twice).** Every pending Linear issue already carries a routing block; all 62 agents are already assigned to 9 lanes. You just couldn't see it. 2. **Config is confirmed cost-safe.** All 62 Paperclip agents + Hermes run gpt-5.5 via Codex/ChatGPT-Pro OAuth — **zero Claude/Anthropic**, no API billing. Your Claude tokens are fully preserved. ✅ 3. **The backlog is real and ~57% agent-doable now** (~46 of ~80 enumerated open issues don't need you). 4. **The gap is activation + visibility, not work.** Everything is *mapped-not-applied* (8 of 9 lanes), the loop isn't running to consume it, and Mission Control shows the roster not the work. ## 1. Hermes already reconciled the backlog (the work-source exists) Two passes, both on the VDS, run as Donna/Hermes CoS: - **Linear Factory Reconciliation** (2026-06-26, `/srv/aina/ops/linear-factory-reconciliation/`): exported all of Linear, cross-checked each issue vs GitHub PRs/branches + repo handoffs + Paperclip runs + Fusion logs, and wrote a `<!-- AINA-CONTROL-RECONCILIATION -->` block into issue **descriptions** with verified-state + evidence links + **execution routing (candidate executor / verifier / safety gate / do-not-redo)**. **767 writebacks, 0 failures.** Surfaced as **AIOPS-226**. So every pending issue already says who should do it, who verifies, and whether it's gated. - **Agent-lane mapping** (2026-06-27, `/srv/repos/aina-paperclip-agent-context/`): all 62 agents → **9 lanes** (growth-media 11, security-privacy-compliance 10, qa-release 8, product-platform 7, content-curriculum 6, executive-governance 6, agentops-tools 6, data-personalization 5, research-intelligence 3). **Status: 56 mapped-not-applied, 6 applied-live-verified (content-curriculum lane only).** ## 2. The roster (62 agents, 3 tiers) CEO **Jessica** → **12 dept heads** → 49 specialists. Dept heads: Gandalf(CoS), Moneypenny(Concierge), Frodo(Release), **Jared(AgentOps, 7 subs)**, **Richard(CTO/Platform, 8 subs)**, **Monica(Product&Learning, 5)**, **Laurie(Data&Personalization, 5)**, **Erlich(Media, 6)**, **Harvey(Growth, 4)**, **Benjamin(Security&Privacy, 4)**, CCO(Compliance, 4 — paused), **Mike(Research&Intel, 3)**. Embedded reviewers in-org: Gimli(code), Calibrator(eval), Silva(security), Editorial Critic, Argus(inspector). All paused/idle, heartbeat off. ## 3. Finch's canon — staged, and not actually blocking 62 per-agent `CONTEXT-MANIFEST.yml` (PKM reference layer, pointer-not-inlined) + 11 dept REFERENCES.md, classified canon(17)/historical/not-canon/stale across 29,351 sources, authority order **Ali → GitHub → Linear → Paperclip → repo docs → PKM wiki → PKM raw → scratch** (PKM is reference *below* execution truth). **Staged, attach_allowed=false** (gated on a fresh non-degraded PKM brain-health + the AIN-47 founder token). **BUT the live agents already carry their base AGENTS.md/DEPARTMENT.md/MEMORY.md bundles** (separate earlier activation, files_ok=224) — so dept heads are NOT context-less; Finch's is an *additional* reference layer, not a prerequisite to run. ## 4. The backlog split (~80 open issues enumerated, ~57% agent-doable) **Agent-doable NOW (~46 — the work the paused fleet is missing):** - Trust-firewall: **ACAD-112** (Urgent, calibration), **ACAD-114** (adaptive-protection 0/20) — *note: these are the redirect-gate in the temporary CLAUDE.md guard; reconcile/route around them first.* - Academy quality loop: ACAD-60/81/97/99/109/110/1, ACAD-69 (QA 225 lessons; founder only on final promote). - AgentOps factory/devex: AIOPS-181(p0 instruction-governance)/192/197/34/33/18/22/184/235/204/206/205, heartbeat + assignee-protocol, **the 13 Paperclip dept-head report ingests (AIN-52..67)**. - Agent Memory ×3, the 6 VDS local-first lanes (MCP/admin/analytics/inbox/PE/video), PLAT-3, AIOPS-tessl/sp3b, VISION-14/16 (design draft, founder-ratify). **Founder-gated (~34 — needs you):** - **Public/beta launch chain** (biggest cluster): PLAT-4/5/6/7, LEARNIN-34, INFRA-42, AIOPS-193 (BETA-LIVE, status:blocked). - Payments/pricing: PLAT-40, VISION-25, ACAD-13. Secrets: AIOPS-35. Supervised Fusion: the data-import mission. - Doctrine/identity/strategy: most of Vision Core (VISION-3/4/5/6/17/18/19/20/23/24), AIOPS-186/182, ACAD-67/120. - Data Engine (DAT): **0 open — fully closed.** ## 5. The SDLC plan (defined, two layers) - **Canonical** (`~/plugins/aina-agentops/`, `references/lifecycle.md`): 8-stage lifecycle (daily-gate → idea → decision-artifact → Linear commitment → GitHub range setup → agent execution → PR+CI → staging proof → prod+retro). Hierarchy: Initiative→Project/Slice→Issue-Range→Issue→Release-Gate. Non-negotiables: 1 issue = 1 branch = 1 PR = 1 builder + 1 reviewer; builders never approve their own gate; no issue = no work. - **Future-state catalog** (`/srv/repos/agentops-for-ali/skills-and-plugins/sdlc-agent-skills-catalog.md`): AINA is strong in the MIDDLE, thin at the EDGES. P0 skills to build: write-spec, author-goal, **dr-gate (done-review)**, decompose-to-beads, **route-from-beads**, prompt-author, self-modification-guard, **start-reconcile**, standards-enforcer, merge-gate. ## 6. The triage + roadmap agent (Ali's ask) **Don't create a new one — operationalize Donna (Hermes CoS).** She already does the reconciliation + lane mapping; the open issue **AIOPS-228 "Enable Donna CoS v1 autonomy without making Ali the bottleneck"** (In Progress, Urgent) IS this. What "triage + roadmap" needs to become a *running loop*: - **Triage** = Donna's reconciliation run on a cadence (daily) → each new/changed issue gets state + executor + gate; the SDLC P0 skills `start-reconcile` + `route-from-beads` + `dr-gate` are the missing skill mechanics. - **Roadmap** = the reconciled backlog + lane assignments, surfaced in Mission Control as the work-board (per the PostHog/Mission-Control plan), with the founder-gated pile as the "What Needs You" rail. - **Actuator** = the Hermes→Paperclip dispatch bridge (already speced) so triage → assignment → run is automatic. Jared's AgentOps team owns the routing mechanics; Donna owns the triage/roadmap decision. ## 7. Codex config — CONFIRMED cost-safe (Ali's check) | Scope | Model | Auth | On Codex | |---|---|---|---| | Paperclip company (+ all 62 subagents inherit) | gpt-5.5, no provider override (OpenAI) | ChatGPT-Pro OAuth, no API key | ✅ | | VDS Hermes (+ 13 aux slots) | gpt-5.5 / openai-codex | ChatGPT OAuth, no Anthropic | ✅ | | Mac Hermes (+ 14 aux slots provider:auto) | gpt-5.5 / openai-codex | ChatGPT OAuth; **no Anthropic credential exists → auto can't route to Claude** | ✅ | `allOnCodex=true, exceptions=[]`. Claude was deliberately stripped 2026-06-28 (`*-remove-claude-anthropic` backups). **Optional belt-and-braces** (not required, zero exposure today): pin the Mac Hermes 14 aux slots `auto→openai-codex` to match the VDS. Lone cleanup: Finch still has `heartbeat.enabled=true` while paused. ## 8. The activation gap + next steps (everything is mapped, nothing runs) Same pattern as the whole org: **excellent staged work, not activated.** To turn the map into a running loop: 1. **Surface it** — Mission Control shows the reconciled backlog + lane assignments + the founder rail (visibility). 2. **Apply the lanes** — 8 of 9 lanes are mapped-not-applied (gives agents scoped cwd/context); reversible, but it moves agent state → founder-gated-ish. Or let the dispatch bridge carry context per-task instead. 3. **Build/activate the bridge** — speced; needs per-agent API keys provisioned (founder gate). 4. **Run Donna's triage on a cadence** — the daily roadmap loop (AIOPS-228). Founder gates remain: Finch attach (optional, agents have base bundles), the public-launch chain, payments, doctrine.
File created successfully at: /Users/Ali/PKM-monorepo/docs/pkm-deepdive/AGENT-WORK-MAP.md (file state is current in your context — no need to Read it back)
cd /Users/Ali/PKM-monorepo cat >> docs/pkm-deepdive/RUNLOG.md <<'EOF' ## 2026-06-29 — Linear→agent mapping investigation DONE (`wwd3qzfgc`): the map already exists - **THE MAP IS ALREADY DONE (twice) — Ali couldn't see it.** Hermes/Donna ran: (1) Linear Factory Reconciliation (2026-06-26, /srv/aina/ops/linear-factory-reconciliation/) → 767 issue-description writebacks (0 fail) each with verified-state + evidence + execution routing (executor/verifier/gate/do-not-redo); surfaced as AIOPS-226. (2) Agent-lane mapping (2026-06-27, /srv/repos/aina-paperclip-agent-context/) → all 62 agents → 9 lanes; 56 mapped-not-applied, 6 applied-live (content-curriculum only). - **Codex config CONFIRMED cost-safe (Ali's check ✅):** all 62 Paperclip agents (gpt-5.5/codex_local, inherit) + VDS Hermes (gpt-5.5/openai-codex, 13 aux pinned) + Mac Hermes (gpt-5.5/openai-codex, 14 aux provider:auto BUT no Anthropic cred so auto can't reach Claude). allOnCodex=true, exceptions=[]. Claude stripped 06-28. Optional pin: Mac aux auto→openai-codex (belt-braces). Finch heartbeat.enabled=true while paused = lone cleanup. - **Backlog ~80 open issues, ~57% agent-doable** (~46 no Ali gate: ACAD quality loop + trust-firewall 112/114, AIOPS factory/devex + 13 dept-report ingests, Agent-Memory ×3, 6 VDS lanes); ~34 founder-gated (BETA-LIVE public launch chain PLAT-4/5/6/7+LEARNIN-34+INFRA-42, payments, doctrine/Vision-Core, secrets). DAT = 0 open (done). - **Finch canon = staged (attach gated) but NOT blocking** — live agents already carry base AGENTS/DEPARTMENT/MEMORY bundles (files_ok=224); Finch's is an additional PKM-reference layer. - **SDLC defined:** canonical 8-stage lifecycle (~/plugins/aina-agentops/) + future-state P0 skills catalog (route-from-beads/start-reconcile/dr-gate = the triage/roadmap mechanics). - **Triage/roadmap agent = operationalize Donna (Hermes CoS), don't build new** — AIOPS-228 IS this. Needs: run reconciliation on a cadence + the dispatch bridge actuator + Mission-Control surfacing. - **THE GAP = activation + visibility, not work.** Same pattern as the whole org: excellent staged work, not run. Synthesis landed: `docs/pkm-deepdive/AGENT-WORK-MAP.md`. - NEXT: surface the map (Mission Control) + decide activation sequence (apply lanes / bridge keys / Donna cadence). EOF git add docs/pkm-deepdive/AGENT-WORK-MAP.md docs/pkm-deepdive/RUNLOG.md git commit -q -m "docs(pkm-deepdive): agent work map — the Linear->agent mapping already exists (staged+invisible) Investigation wwd3qzfgc: Hermes already reconciled the backlog (767 issue routing writebacks, 0 fail) + mapped all 62 agents to 9 lanes (56 mapped-not-applied). Config confirmed cost-safe (all gpt-5.5/ codex, zero Claude). ~57% of the ~80 open issues are agent-doable now. Finch canon staged but not blocking (agents have base bundles). Triage/roadmap = operationalize Donna (AIOPS-228), don't rebuild. Gap = activation + visibility, not work. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>" git push -q origin main && echo PUSHED
Mapped, confirmed, and landed (AGENT-WORK-MAP.md). Here's the picture — and it should make you feel better about how much is actually done:
{
"title": "agent_work_map_backlog_split",
"loading_messages": [
"Sorting the backlog",
"Splitting agent-doable from gated",
"Checking who's actually running",
"Tallying what needs you"
],
"widget_code": "<svg viewBox=\"0 0 760 500\" xmlns=\"http://www.w3.org/2000/svg\" role=\"img\" font-family=\"ui-sans-serif, system-ui, -apple-system, Segoe UI, Roboto, sans-serif\">\n <title>Agent work map: the backlog is mapped and config is cost-safe, but the agents aren't running it</title>\n <desc>Of ~80 open Linear issues, ~46 are agent-doable now and ~34 need the founder. All 767 issues and 62 agents are already mapped to 9 lanes, but only 1 lane is live, so nothing runs. Config is confirmed all-Codex, Claude tokens safe.</desc>\n <style>\n .h{fill:var(--text-primary,#1a1a1a);font-weight:750;font-size:14px}\n .lbl{fill:var(--text-primary,#1a1a1a);font-weight:650;font-size:12.5px}\n .sub{fill:var(--text-secondary,#6b7280);font-size:10.5px}\n .tiny{fill:var(--text-secondary,#6b7280);font-size:10px}\n .ready{fill:var(--accent-teal-bg,#e7f7f1);stroke:var(--accent-teal,#0f9d76);stroke-width:1.5}\n .gated{fill:var(--accent-amber-bg,#fdf3e3);stroke:var(--accent-amber,#d8902b);stroke-width:1.5}\n .chipok{fill:var(--accent-teal-bg,#e7f7f1);stroke:var(--accent-teal,#0f9d76);stroke-width:1.3}\n .chipno{fill:var(--accent-red-bg,#fdecec);stroke:var(--accent-red,#d6453d);stroke-width:1.3}\n .chiptxt{fill:var(--text-primary,#1a1a1a);font-size:11px;font-weight:600}\n .big{fill:var(--text-primary,#1a1a1a);font-weight:800;font-size:26px}\n </style>\n\n <!-- status strip -->\n <text x=\"20\" y=\"26\" class=\"h\">Where the work stands</text>\n <g>\n <rect x=\"20\" y=\"38\" width=\"232\" height=\"34\" rx=\"8\" class=\"chipok\"/>\n <text x=\"34\" y=\"53\" class=\"chiptxt\">✓ Mapped</text>\n <text x=\"34\" y=\"67\" class=\"tiny\">767 issues + 62 agents → 9 lanes</text>\n\n <rect x=\"264\" y=\"38\" width=\"232\" height=\"34\" rx=\"8\" class=\"chipno\"/>\n <text x=\"278\" y=\"53\" class=\"chiptxt\">✗ Running</text>\n <text x=\"278\" y=\"67\" class=\"tiny\">only 1 of 9 lanes live · fleet paused</text>\n\n <rect x=\"508\" y=\"38\" width=\"232\" height=\"34\" rx=\"8\" class=\"chipok\"/>\n <text x=\"522\" y=\"53\" class=\"chiptxt\">✓ Config cost-safe</text>\n <text x=\"522\" y=\"67\" class=\"tiny\">all on Codex/gpt-5.5 · Claude safe</text>\n </g>\n\n <!-- two columns -->\n <!-- LEFT: agent-doable -->\n <rect x=\"20\" y=\"92\" width=\"355\" height=\"320\" rx=\"11\" class=\"ready\"/>\n <text x=\"38\" y=\"120\" class=\"big\">~46</text>\n <text x=\"120\" y=\"112\" class=\"lbl\">Agents can start now</text>\n <text x=\"120\" y=\"128\" class=\"sub\">no gate from you · already routed to lanes</text>\n <text x=\"38\" y=\"158\" class=\"lbl\">Academy product-quality loop</text>\n <text x=\"38\" y=\"174\" class=\"tiny\">engine wiring, evaluators, 225-lesson QA, trust-firewall 112/114</text>\n <text x=\"38\" y=\"198\" class=\"lbl\">AgentOps factory + dev-ex</text>\n <text x=\"38\" y=\"214\" class=\"tiny\">instruction governance (p0), branch cleanup, VDS parity, smokes</text>\n <text x=\"38\" y=\"238\" class=\"lbl\">Ingest the 13 dept-head reports</text>\n <text x=\"38\" y=\"254\" class=\"tiny\">the work the agents already produced → into Linear</text>\n <text x=\"38\" y=\"278\" class=\"lbl\">Agent Memory ×3 + 6 VDS lanes</text>\n <text x=\"38\" y=\"294\" class=\"tiny\">PKM access model, MCP/admin/analytics/inbox/PE/video builds</text>\n <text x=\"38\" y=\"318\" class=\"lbl\">Front-end cleanup, Codex-fix, design drafts</text>\n <text x=\"38\" y=\"334\" class=\"tiny\">PLAT-3, sp3b, Vision research (you ratify, agents draft)</text>\n <line x1=\"38\" y1=\"356\" x2=\"357\" y2=\"356\" stroke=\"var(--border,#d6dae0)\" stroke-width=\"1\"/>\n <text x=\"38\" y=\"378\" class=\"sub\">Lanes ready: product-platform · agentops-tools · qa-release ·</text>\n <text x=\"38\" y=\"393\" class=\"sub\">data-personalization · research-intelligence</text>\n\n <!-- RIGHT: founder-gated -->\n <rect x=\"387\" y=\"92\" width=\"353\" height=\"320\" rx=\"11\" class=\"gated\"/>\n <text x=\"405\" y=\"120\" class=\"big\">~34</text>\n <text x=\"487\" y=\"112\" class=\"lbl\">Needs you</text>\n <text x=\"487\" y=\"128\" class=\"sub\">decisions, money, launch, secrets</text>\n <text x=\"405\" y=\"158\" class=\"lbl\">Public / beta launch chain</text>\n <text x=\"405\" y=\"174\" class=\"tiny\">BETA-LIVE-00..09 · the biggest gated cluster · status:blocked</text>\n <text x=\"405\" y=\"198\" class=\"lbl\">Payments + pricing</text>\n <text x=\"405\" y=\"214\" class=\"tiny\">Dodo application, capability pricing, Lab-OS cost analysis</text>\n <text x=\"405\" y=\"238\" class=\"lbl\">Doctrine / identity / strategy</text>\n <text x=\"405\" y=\"254\" class=\"tiny\">most of Vision Core — manifesto, constitution, positioning</text>\n <text x=\"405\" y=\"278\" class=\"lbl\">Secrets + supervised runs</text>\n <text x=\"405\" y=\"294\" class=\"tiny\">observability credentials, the supervised Fusion data-import</text>\n <line x1=\"405\" y1=\"316\" x2=\"723\" y2=\"316\" stroke=\"var(--border,#d6dae0)\" stroke-width=\"1\"/>\n <text x=\"405\" y=\"338\" class=\"sub\">These become your \"What Needs You\" rail —</text>\n <text x=\"405\" y=\"353\" class=\"sub\">the loop surfaces them to you, doesn't wait on you for the rest.</text>\n\n <!-- takeaway -->\n <rect x=\"20\" y=\"426\" width=\"720\" height=\"56\" rx=\"10\" fill=\"var(--bg-surface,#f6f7f9)\" stroke=\"var(--border,#d6dae0)\" stroke-width=\"1.5\"/>\n <text x=\"380\" y=\"450\" text-anchor=\"middle\" class=\"h\">The work is mapped. The agents are paused. The config is safe.</text>\n <text x=\"380\" y=\"470\" text-anchor=\"middle\" class=\"sub\">The gap is activation + visibility — not work, and not you being the bottleneck on ~46 of these.</text>\n</svg>"
}Content rendered and shown to the user. Please do not duplicate the shown content in text because it's already visually represented.
Two things the picture doesn't say out loud:
Both your "for now" asks are essentially already done — you just couldn't see them. You asked Hermes to reconcile the backlog so it could be assigned to agents; she did, on June 26: 767 issues each got an evidence-backed routing block (who executes, who verifies, whether it's gated), zero failures, and a follow-up pass put all 62 agents into 9 lanes. And your codex-config check is confirmed clean — every agent and Hermes runs gpt-5.5 on your ChatGPT/Codex subscription, zero Claude, no API billing. Claude tokens fully protected. (One tiny cleanup: Finch still has its heartbeat flag on while paused.)
Your "triage and roadmap agent" already exists — it's Donna. Don't build a new one. Hermes-as-Donna is the Chief of Staff that reconciled and routed everything, and there's even an open issue for exactly this goal: AIOPS-228, "Enable Donna CoS v1 autonomy without making Ali the bottleneck." What's missing isn't the agent — it's making her reconciliation run on a cadence (the daily roadmap) and giving her the dispatch bridge to actuate it. The SDLC plan even names the exact skills to build for it (route-from-beads, start-reconcile, dr-gate).
And the canon isn't actually blocking you. Finch's manifests are staged, yes — but the live agents already carry their base instruction bundles (224 files attached in an earlier pass). Finch's layer is an upgrade, not a prerequisite. So that founder gate isn't on the critical path right now.
So the honest state: the map is built, the config is safe, ~46 issues are ready for agents that aren't running, and ~34 are genuinely yours. The gap — every single time — is activation and visibility, not work, and not you.
Given that, what do you want me to turn into a running loop first?
{
"questions": [
{
"question": "Both 'for now' asks are done (backlog mapped + config confirmed cost-safe). What should I turn into a running loop next? (Agents stay paused until you explicitly activate.)",
"header": "Next move",
"multiSelect": false,
"options": [
{
"label": "Surface the map in Mission Control (Recommended)",
"description": "Make the cockpit show the reconciled backlog: the ~46 ready issues + their routing/lanes, and the ~34 that need you as a 'What Needs You' rail. You finally SEE the work + what's blocked on you, in one place. Safe, nothing wakes — visibility first, which everything else builds on."
},
{
"label": "Stand up Donna's triage/roadmap as a daily loop",
"description": "Operationalize AIOPS-228: run Donna's reconciliation on a cadence so the backlog stays auto-routed + the roadmap self-updates. The 'keeps moving without me' engine. Builds the route/reconcile/done-review skills."
},
{
"label": "Build the dispatch bridge + provision keys",
"description": "The actuator that lets Donna actually start a dept-head on a ready issue. Founder-gated: provisioning per-agent API keys + the first live (non-dry) runs mean agents run. Highest leverage toward autonomy, biggest step."
}
]
}
]
}Your questions have been answered: "Both 'for now' asks are done (backlog mapped + config confirmed cost-safe). What should I turn into a running loop next? (Agents stay paused until you explicitly activate.)"="Surface the map in Mission Control (Recommended)". You can now continue with these answers in mind.
Surface the map it is — visibility first, nothing wakes. That means building a "Work Map" view into Mission Control: the ~46 ready issues + their lane routing, and the ~34 that need you as a rail. Before I touch the cockpit (it's your live auth-gated surface), let me reconcile exactly how it's structured + how it gets data, so I build it the right way.
{
"title": "Surface work-map in Mission Control",
"summary": "Build a Work Map view into the existing Mission Control cockpit: reconciled backlog (agent-doable vs founder-gated) + lane routing, reading from a work-map data artifact. Reconcile the repo + data pipeline first."
}cd ~ echo "=== locate Mission Control repo(s) ===" for d in ~/Projects/aina-mission-control ~/aina-mission-control ~/conductor/repos/aina-mission-control; do [ -d "$d" ] && echo "FOUND: $d (branch: $(git -C "$d" branch --show-current 2>/dev/null), last: $(git -C "$d" log -1 --format='%h %cr' 2>/dev/null))" done echo R=~/Projects/aina-mission-control [ -d "$R" ] || R=~/aina-mission-control echo "=== using R=$R ; top-level structure ===" ls -1 "$R" 2>/dev/null | head -30 echo echo "=== functions/api routes (the Worker endpoints that serve cards) ===" find "$R/functions" -type f 2>/dev/null | head -40 echo echo "=== focus feed + collectors + wrangler (data pipeline) ===" find "$R" -maxdepth 3 \( -name 'wrangler.toml' -o -name '*focus*' -o -name '*collector*' -o -path '*collector*' -o -name 'feed*.json' \) -not -path '*/node_modules/*' 2>/dev/null | head -30
=== locate Mission Control repo(s) === FOUND: /Users/Ali/Projects/aina-mission-control (branch: main, last: f835702 2 days ago) === using R=/Users/Ali/Projects/aina-mission-control ; top-level structure === DESIGN.md README.md RUNLOG.md STATE.md collector deploy.sh design dist docs functions index.html node_modules ops package-lock.json package.json postcss.config.js public push-worker src tailwind.config.js test test-py tsconfig.json vite.config.ts vitest.config.ts wrangler.toml === functions/api routes (the Worker endpoints that serve cards) === /Users/Ali/Projects/aina-mission-control/functions/_middleware.js /Users/Ali/Projects/aina-mission-control/functions/api/linear.js /Users/Ali/Projects/aina-mission-control/functions/api/health.js /Users/Ali/Projects/aina-mission-control/functions/api/factory.js /Users/Ali/Projects/aina-mission-control/functions/api/agent-ops.js /Users/Ali/Projects/aina-mission-control/functions/api/delivery.js /Users/Ali/Projects/aina-mission-control/functions/api/marketing.js /Users/Ali/Projects/aina-mission-control/functions/api/readiness.js /Users/Ali/Projects/aina-mission-control/functions/api/github.js /Users/Ali/Projects/aina-mission-control/functions/api/infra.js /Users/Ali/Projects/aina-mission-control/functions/api/analytics.js /Users/Ali/Projects/aina-mission-control/functions/api/errors.js /Users/Ali/Projects/aina-mission-control/functions/api/product.js /Users/Ali/Projects/aina-mission-control/functions/_shared/cards.js /Users/Ali/Projects/aina-mission-control/functions/_shared/factory.js /Users/Ali/Projects/aina-mission-control/functions/_shared/feed.js /Users/Ali/Projects/aina-mission-control/functions/_shared/series.js /Users/Ali/Projects/aina-mission-control/functions/_shared/activity.js /Users/Ali/Projects/aina-mission-control/functions/_shared/delivery.js /Users/Ali/Projects/aina-mission-control/functions/_shared/freshness.js /Users/Ali/Projects/aina-mission-control/functions/_shared/watchdog.js /Users/Ali/Projects/aina-mission-control/functions/_shared/sentry.js /Users/Ali/Projects/aina-mission-control/functions/_shared/product.js /Users/Ali/Projects/aina-mission-control/functions/_shared/http.js /Users/Ali/Projects/aina-mission-control/functions/api/pkm/activity.js /Users/Ali/Projects/aina-mission-control/functions/api/focus/feed.json.js /Users/Ali/Projects/aina-mission-control/functions/api/push/digest.js === focus feed + collectors + wrangler (data pipeline) === /Users/Ali/Projects/aina-mission-control/collector /Users/Ali/Projects/aina-mission-control/wrangler.toml /Users/Ali/Projects/aina-mission-control/push-worker/wrangler.toml /Users/Ali/Projects/aina-mission-control/collector/repo_freshness.py /Users/Ali/Projects/aina-mission-control/collector/paperclip_snapshot.py /Users/Ali/Projects/aina-mission-control/collector/migrations /Users/Ali/Projects/aina-mission-control/collector/collect.py /Users/Ali/Projects/aina-mission-control/collector/schema.sql /Users/Ali/Projects/aina-mission-control/collector/deploy /Users/Ali/Projects/aina-mission-control/collector/dedupe.py /Users/Ali/Projects/aina-mission-control/collector/__pycache__ /Users/Ali/Projects/aina-mission-control/collector/vds_metrics.py /Users/Ali/Projects/aina-mission-control/collector/upsert.py /Users/Ali/Projects/aina-mission-control/collector/migrations/0002_paperclip.sql /Users/Ali/Projects/aina-mission-control/collector/deploy/com.ali.mission-collector.plist /Users/Ali/Projects/aina-mission-control/collector/deploy/mission-collector.service /Users/Ali/Projects/aina-mission-control/collector/deploy/mission-collector.timer /Users/Ali/Projects/aina-mission-control/collector/__pycache__/upsert.cpython-37.pyc /Users/Ali/Projects/aina-mission-control/collector/__pycache__/upsert.cpython-312.pyc /Users/Ali/Projects/aina-mission-control/collector/__pycache__/dedupe.cpython-37.pyc /Users/Ali/Projects/aina-mission-control/collector/__pycache__/vds_metrics.cpython-312.pyc /Users/Ali/Projects/aina-mission-control/collector/__pycache__/dedupe.cpython-312.pyc /Users/Ali/Projects/aina-mission-control/functions/api/focus Shell cwd was reset to /Users/Ali/PKM
Solid — Mission Control is a Cloudflare Pages + Functions app with a clean pattern: Python collectors → D1/R2 → functions/api/*.js Workers → frontend cards. Crucially, functions/api/linear.js already exists, so there's a Linear data path to build on. Let me read the repo's own state doc + the existing Linear endpoint + the frontend card pattern, so the Work Map slots into the existing architecture instead of fighting it:
1 # STATE — AINA Mission Control ("The Prancing Pony")
2
3 **Updated:** 2026-06-27 · **Live:** https://mission.ainative.academy (Basic Auth, fail-closed)
4 **Linear:** AIOPS-183 (v0, In Review) · AIOPS-216 (v1 redesign) · AIOPS-215 (integration spine) · **Repo:** `ainative-academy/aina-mission-control`
5
6
7 > **2026-06-27 — Founder/friends -> paid-beta -> public runtime gates drafted.** Mission Control now has a founder-readable gate spec for the staged path Ali chose: first private founder/friends proof, then discount-code paid beta, then public launch unlock. The static readiness fallback now treats academy PR #73 / ACAD-134 as GitHub-cleared but Linear-stale, while preserving `not_live_ready` / `cohort_blocked` because PLAT-11/12, DAT-91, privacy/legal receipts, process-quality proof, and paid-beta commerce gates remain locked.
8
9 > **2026-06-26 — Readiness control pane landed and deployed.** Mission Control now has a Readiness tab, Bridge readiness panel, waiting-on-Ali rail, read-only `/api/readiness` contract, Paperclip lane mapping overlay, and read-only VDS repo freshness collector. Verification on the VDS: `npm test` = 56 tests green, `npm run build` green, `python3 collector/repo_freshness.py --pretty` green. Claude CLI copilot review was attempted but unavailable on the VDS due `401 Invalid authentication credentials`; Codex adversarial review then found static-freshness, liveness-dot, Paperclip-advisory, optional-git-lock, and untracked-file risks, all fixed before commit. This checkpoint remains not_live_ready/cohort_blocked by design; no Paperclip/RunFusion resume and no public/launch gate changes.
10
11 > **2026-06-26 — AINA live-readiness control-pane pivot.** Phase 2 data spine is complete and the site is Basic-Auth gated and live. Mission Control is now the founder-facing readiness pane for AINA: academy runtime truth, DER supplier blockers, platform/control-plane PRs, Paperclip roster, and waiting-on-Ali gates. Current reconciliation remains intentionally not_live_ready/cohort_blocked: academy PR #73 is parked under ACAD-134, DER still carries DAT-94/DAT-96/DAT-97 supplier blockers, platform remains reference/control-plane unless founder cutover changes it, and production/public gates stay founder-held. Plan: docs/plans/2026-06-26-001-feat-aina-live-readiness-control-pane-plan.md.
12
13 > **2026-06-25 — redesign + integration scope.** Design-shotgun done (4 variants in `design/mockups/`, preview at `/design/`); direction **LOCKED = Variant D "AINA Luminous"** (AINA Design System / luminous-glass reskin), approved by Ali. Live Paperclip/RunFusion/GitHub integration mapped + proven on the VDS. Full detail: **`docs/handoff/2026-06-25-redesign-and-integration-map.md`**. Sequence: paint first → then data.
14 > - **✅ Lane A LANDED (port `main` @ `d31b3dd`, deployed):** live cockpit re-themed to Luminous Glass — navy ambient backdrop, teal/gold, Playfair+Inter+JetBrains Mono, glass shell/components. Routes/nav/data UNCHANGED (theme+shell only). 29 tests green. Branch `ali/luminous-port-2026-06-25`.
15 > - **✅ Lane B LANDED (`main` @ `f731917`, deployed 2026-06-26 from the VDS):** the dense command-center **Bridge** layout — 4-tab IA (BRIDGE/FACTORY/SYSTEMS/OPS), Bridge home (focal "What Needs You" from api.feed + speedometer + gauge rings from agentOps/github/linear + Factory/Systems/Delivery panels + comms log from api.activity), honest not-wired for Paperclip/RunFusion/Sentry. **Built by Codex on the VDS** (via Cloudflare Access `aina-vds-cf`; Tailscale was stopped), **Codex second-opinion review caught 5 honesty/liveness bugs → Sonnet-fixed**. 29 tests green; read-only intact. Branch `ali/luminous-bridge-2026-06-26`.
16 > - **✅ Phase 2 — data spine COMPLETE** (AIOPS-215). LANDED + deployed: Paperclip, GitHub Delivery/CI, Cloudflare traffic, Sentry, PostHog Product, and Slack daily push. RunFusion deferred (daemon paused). Detail: docs/handoff/2026-06-26-phase2-progress.md + RUNLOG.md.
17
18 ## Current state (deployed)
19 Production = readiness control pane deployed from origin/main at `2b413e7` (Cloudflare Pages deploy `327c76e8`). Founder ops surface: **Bridge / Factory / Readiness / Systems / Ops** command-center (Luminous Glass), read-only, live-data-only where connectors exist, Basic-Auth gated, and explicitly **not_live_ready/cohort_blocked** for AINA launch claims. Current local change drafts the founder/friends runtime gates and updates static readiness guardrails, but has not been deployed yet. Live or intentionally guarded data: Paperclip · GitHub delivery · Cloudflare · Sentry · PostHog · Slack daily push · VDS metrics · Linear · agent-ops · activity · `/api/readiness` static guardrails. *(VDS reached via aina-vds-cf when Tailscale is down.)*
20
21 **Data flowing (verified live 2026-06-24):**
22 - Mac collector (launchd, every 15 min): `OK host=mac sessions=108 reconciliation_pushed=True`.
23 - VDS collector (systemd, every 15 min): `OK host=vds metrics=4` → `vds.cpu/mem/disk/docker` fresh in MISSION_DB.
24 - Connectors live: Linear, GitHub (Pages secrets set), PKM_DB / MISSION_DB / R2.
25
26 ## Shipped / active control-pane truth
27 - **Readiness pane shipped** — Bridge readiness panel, full Readiness tab, Mission/Milestone/Slice/Task tree, waiting-on-Ali rail, and static `/api/readiness` guardrails.
28 - **Phase 2 data spine complete** — Paperclip, GitHub Delivery/CI, Cloudflare traffic, Sentry, PostHog Product, Slack daily push, VDS metrics, Linear, agent-ops, and activity are represented without launch overclaiming.
29 - **Reviewer reconciliation active** — current follow-up fixes stale `STATE.md` text, missing `/api/readiness` issue rows, and production-broken repo-doc provenance links from Codex reviewer comments on PRs #2-#4.
30 - **RunFusion deferred** — daemon remains paused; Mission Control may show advisory routing, but does not resume Paperclip/RunFusion or execute public/live gates.
31
32 ## Founder-gated / intentionally not changed here
33 1. **AINA live/cohort readiness** — still blocked by academy proof, telemetry/privacy ordering, DER supplier prerequisites, and founder gates.
34 2. **Public apex / production unlocks** — PLAT-11/12 and DAT-91 remain founder-held.
35 3. **Payments, billing, sensitive-role runtime, process-quality mastery promotion** — not executable from Mission Control.
36 4. **Analytics/Marketing expansion** — `docs/connectors.md` remains the activation map for optional deeper connectors.
37 5. **Live screenshot proof** — cockpit is Basic-Auth gated; rendered proof still needs founder-held access credentials.
38
39 ## Deploy note (Mac network blip 2026-06-24)
40 The Mac could not reach `dash.cloudflare.com` via Node/undici (curl worked — a Node happy-eyeballs stall). **Deployed from the VDS instead.** Resume deploy:
41 ```bash
42 # Mac (if network OK): cd ~/Projects/aina-mission-control && ./deploy.sh
43 # VDS fallback:
44 ssh aina-vds-ts 'cd /home/ali/aina-mission-control && git fetch origin -q && git reset --hard origin/main && npm ci && npm run build && CLOUDFLARE_ACCOUNT_ID=1013214185b301441df8053a8c594275 wrangler pages deploy dist --project-name=aina-mission-control'
45 ```
46
47 ## Not in this build (deferred / next program)
48 Metric-history snapshots (so Linear/GitHub/VDS count cards trend over time) · AIOPS-215 event spine · AIOPS-216 Sentry/PostHog dashboard-v1 · AIOPS-54 Linear-independence.
49
50 ## Tests
51 Latest readiness checkpoint: `npm test` (root, 58) · `npm run build`. Historical Phase 2 checks also covered `python3 collector/repo_freshness.py --pretty` on the VDS, `cd push-worker && npm test` (10), and `python3 -m unittest discover -s test-py` (14).
52 R=/Users/Ali/Projects/aina-mission-control echo "=== functions/api/linear.js (head: auth + data source) ===" sed -n '1,60p' "$R/functions/api/linear.js" echo echo "=== how the frontend defines tabs/cards (src/) ===" ls -1 "$R/src" 2>/dev/null | head -30 echo "--- grep tab/card definitions ---" grep -rnE "READINESS|FACTORY|BRIDGE|tabs?\s*[:=]|TABS|card" "$R/src" 2>/dev/null | grep -iE "tab|nav" | head -15
=== functions/api/linear.js (head: auth + data source) ===
// /api/linear — near-live Linear delivery cards via the Linear GraphQL API
// (fetch, read-only). No token -> "not wired" (never a fake number).
import { makeCard, notWired, staleFrom } from "../_shared/cards.js";
import { TARGETS } from "../_shared/freshness.js";
import { json } from "../_shared/http.js";
const PAGE_CAP = 20;
async function countIssues(token, filter) {
let total = 0;
let after = null;
let hasNextPage = true;
let page = 0;
while (hasNextPage && page < PAGE_CAP) {
const variables = after ? { f: filter, after } : { f: filter };
const query = after
? `query($f: IssueFilter, $after: String){ issues(filter:$f, after:$after){ pageInfo{ hasNextPage endCursor } nodes{ id } } }`
: `query($f: IssueFilter){ issues(filter:$f){ pageInfo{ hasNextPage endCursor } nodes{ id } } }`;
const res = await fetch("https://api.linear.app/graphql", {
method: "POST",
headers: { Authorization: token, "content-type": "application/json" },
body: JSON.stringify({ query, variables }),
});
if (!res.ok) throw new Error(`linear ${res.status}`);
const data = await res.json();
const issues = data?.data?.issues;
const nodes = issues?.nodes ?? [];
total += nodes.length;
hasNextPage = issues?.pageInfo?.hasNextPage ?? false;
after = issues?.pageInfo?.endCursor ?? null;
page++;
}
return total;
}
export async function onRequestGet({ env }) {
const token = env.LINEAR_API_TOKEN;
const now = Date.now();
const url = "https://linear.app";
if (!token) {
return json({ available: false, cards: [
notWired({ key: "linear.blocked", label: "Linear blocked", source: "Linear", drillUrl: url }),
notWired({ key: "linear.urgent", label: "Linear urgent", source: "Linear", drillUrl: url }),
] });
}
const card = (key, label, value) =>
makeCard({ key, label, value, source: "Linear", lastRefreshed: new Date(now).toISOString(), drillUrl: url, targetSeconds: TARGETS.linear, nowMs: now });
try {
const [blocked, urgent] = await Promise.all([
// Linear has no "blocked" workflow-state TYPE (enum: triage/backlog/unstarted/
// started/completed/canceled); blocked is a label convention (e.g. status:blocked).
countIssues(token, { labels: { some: { name: { containsIgnoreCase: "blocked" } } } }).catch(() => null),
countIssues(token, { priority: { eq: 1 } }).catch(() => null),
]);
const cards = [];
cards.push(blocked == null ? staleFrom(card("linear.blocked", "Linear blocked", null)) : card("linear.blocked", "Linear blocked", blocked));
cards.push(urgent == null ? staleFrom(card("linear.urgent", "Linear urgent", null)) : card("linear.urgent", "Linear urgent", urgent));
return json({ available: true, cards });
=== how the frontend defines tabs/cards (src/) ===
app.tsx
components
data
dev
index.css
lib
main.tsx
nav.ts
pages
vite-env.d.ts
--- grep tab/card definitions ---Now I understand the cockpit well — React frontend + Worker endpoints + collectors, with api/linear.js already querying Linear live (it only counts blocked/urgent today). Important: STATE.md says there's an undeployed local change in this repo, so before I add anything I must check its git state — I won't clobber in-flight founder-gate work, and I'll build on a branch (this is your live cockpit, so I won't auto-deploy).
R=/Users/Ali/Projects/aina-mission-control echo "=== git state (any uncommitted founder-gate work to preserve?) ===" git -C "$R" status -sb 2>&1 | head -20 echo "--- ahead/behind origin ---" git -C "$R" rev-list --count origin/main..main 2>/dev/null | sed 's/^/ local ahead of origin: /' git -C "$R" rev-list --count main..origin/main 2>/dev/null | sed 's/^/ local behind origin: /' echo echo "=== nav/tabs definition ===" sed -n '1,60p' "$R/src/nav.ts" echo echo "=== pages (each tab) ===" ls -1 "$R/src/pages" 2>/dev/null
=== git state (any uncommitted founder-gate work to preserve?) ===
## main...origin/main
?? .claude/
--- ahead/behind origin ---
local ahead of origin: 0
local behind origin: 0
=== nav/tabs definition ===
// Single source of truth for command-center navigation. Secondary source
// surfaces remain routeable drill-downs, but the primary cockpit IA is the top tab set.
export interface NavEntry {
path: string;
label: string;
}
export const NAV: NavEntry[] = [
{ path: "/", label: "Bridge" },
{ path: "/factory", label: "Factory" },
{ path: "/readiness", label: "Readiness" },
{ path: "/systems", label: "Systems" },
{ path: "/ops", label: "Ops" },
];
=== pages (each tab) ===
Activity.tsx
Analytics.tsx
Bridge.tsx
Factory.tsx
Focus.tsx
Marketing.tsx
Ops.tsx
Readiness.tsx
Systems.tsxMC repo is clean (local == origin, the only untracked is a harmless .claude/ dir — the 06-27 "undeployed change" has since reconciled). 5 tabs in nav.ts, page-per-tab in src/pages/. So the Work Map is a clean ~4-file add (endpoint + page + nav entry + test) on the existing pattern — built on a branch, deployed only on your nod. Reading the page + routing + card conventions so it matches exactly:
1 import { api } from "../lib/api";
2 import { useAsync } from "../lib/useAsync";
3 import { fmtNum } from "../lib/format";
4 import { FactoryCard, GaugeRing, SectionHead } from "../components/command-center";
5 import type { Card } from "../lib/types";
6
7 function byKey(cards: Card[] | undefined, key: string) {
8 return cards?.find((card) => card.key === key);
9 }
10
11 function cardText(card: Card | undefined) {
12 return card ? fmtNum(card.value) : "not wired";
13 }
14
15 function timeText(value: Card["value"] | undefined) {
16 if (typeof value !== "string") return "none";
17 const t = Date.parse(value);
18 return Number.isNaN(t) ? value : new Date(t).toLocaleString();
19 }
20
21 export default function Factory() {
22 const agentOps = useAsync(() => api.agentOps(), []);
23 const factory = useAsync(() => api.factory(), []);
24 const runs = byKey(agentOps.data?.cards, "agentops.runs");
25 const tokens = byKey(agentOps.data?.cards, "agentops.tokens");
26 const paperclipAgents = byKey(factory.data?.cards, "factory.paperclip.agents");
27 const paperclipActive = byKey(factory.data?.cards, "factory.paperclip.active");
28 const paperclipPaused = byKey(factory.data?.cards, "factory.paperclip.paused");
29 const paperclipStuck = byKey(factory.data?.cards, "factory.paperclip.stuck");
30 const paperclipSpend = byKey(factory.data?.cards, "factory.paperclip.spend");
31 const paperclipHeartbeat = byKey(factory.data?.cards, "factory.paperclip.heartbeat");
32 const runValue = typeof runs?.value === "number" ? runs.value : 0;
33 const tokenValue = typeof tokens?.value === "number" ? tokens.value : 0;
34 const paperclipValue = typeof paperclipAgents?.value === "number" ? paperclipAgents.value : 0;
35 const tokenProgress = tokens?.state === "live" ? Math.min(1, tokenValue / 500_000_000) : 0;
36
37 return (
38 <div className="command-center">
39 <section>
40 <SectionHead title="Factory" meta="Agents + orchestration" state={paperclipAgents?.state ?? "not-wired"} />
41 <div className="row-5">
42 <GaugeRing label="Agent runs 7d" value={fmtNum(runs?.value ?? null)} progress={Math.min(1, runValue / 20)} state={runs?.state ?? "stale"} />
43 <GaugeRing label="Tokens 7d" value={fmtNum(tokens?.value ?? null)} progress={tokenProgress} state={tokens?.state ?? "stale"} />
44 <GaugeRing label="Paperclip" value={cardText(paperclipAgents)} progress={paperclipAgents?.state === "live" ? Math.min(1, paperclipValue / 80) : 0} state={paperclipAgents?.state ?? "not-wired"} />
45 <GaugeRing label="RunFusion" value="not wired" progress={0} state="not-wired" />
46 <GaugeRing label="Hermes" value="not wired" progress={0} state="not-wired" />
47 </div>
48 </section>
49
50 <section>
51 <SectionHead title="Factory Health" meta="Paperclip live snapshot" state={paperclipAgents?.state ?? "not-wired"} />
52 <div className="factory-grid">
53 <FactoryCard
54 name="Paperclip"
55 tone={paperclipAgents?.state === "live" ? "go" : "caution"}
56 state={paperclipAgents?.state ?? "not-wired"}
57 rows={[
58 { label: "Agents", value: cardText(paperclipAgents), tone: paperclipAgents?.state === "live" ? "good" : "mid" },
59 { label: "Active / paused / stuck", value: `${cardText(paperclipActive)} / ${cardText(paperclipPaused)} / ${cardText(paperclipStuck)}`, tone: "teal" },
60 { label: "Spend", value: cardText(paperclipSpend), tone: "mid" },
61 { label: "Heartbeat", value: timeText(paperclipHeartbeat?.value) },
62 ]}
63 />
64 <FactoryCard
65 name="Agent Ops"
66 tone={runs?.state === "live" ? "go" : "caution"}
67 state={runs?.state ?? "stale"}
68 rows={[
69 { label: "Runs 7d", value: fmtNum(runs?.value ?? null), tone: "good" },
70 { label: "Tokens 7d", value: fmtNum(tokens?.value ?? null), tone: "teal" },
71 { label: "Source", value: runs?.source ?? "Agent Ops" },
72 ]}
73 />
74 {["RunFusion", "Hermes"].map((name) => (
75 <FactoryCard
76 key={name}
77 name={name}
78 tone="caution"
79 state="not-wired"
80 rows={[
81 { label: "Status", value: "not wired", tone: "mid" },
82 { label: "Phase", value: "Phase 2" },
83 { label: "Truth", value: "planned" },
84 ]}
85 />
86 ))}
87 </div>
88 </section>
89 </div>
90 );
91 }
92 R=/Users/Ali/Projects/aina-mission-control echo "=== app.tsx routing (how pages are wired) ===" grep -nE "Route|path|import.*pages|Factory|Readiness|element" "$R/src/app.tsx" | head -30 echo echo "=== _shared/cards.js (makeCard/notWired helpers) ===" sed -n '1,55p' "$R/functions/_shared/cards.js" echo echo "=== an existing api test (to mirror) ===" ls "$R/test" 2>/dev/null; echo "---"; find "$R/test" -name '*.test.*' 2>/dev/null | head
=== app.tsx routing (how pages are wired) ===
1:import { NavLink, Route, Routes } from "react-router-dom";
8:import Bridge from "./pages/Bridge";
9:import Factory from "./pages/Factory";
10:import Ops from "./pages/Ops";
11:import Activity from "./pages/Activity";
12:import Systems from "./pages/Systems";
13:import Analytics from "./pages/Analytics";
14:import Marketing from "./pages/Marketing";
15:import Readiness from "./pages/Readiness";
144: key={n.path}
145: to={n.path}
146: end={n.path === "/"}
172: <Routes>
173: <Route path="/" element={<Bridge />} />
174: <Route path="/factory" element={<Factory />} />
175: <Route path="/readiness" element={<Readiness />} />
176: <Route path="/activity" element={<Activity />} />
177: <Route path="/systems" element={<Systems />} />
178: <Route path="/ops" element={<Ops />} />
179: <Route path="/analytics" element={<Analytics />} />
180: <Route path="/marketing" element={<Marketing />} />
181: <Route path="*" element={<div className="font-mono text-sm text-ink-faint">Not found.</div>} />
182: </Routes>
=== _shared/cards.js (makeCard/notWired helpers) ===
// Card response contract (R5) + the live / stale / not-wired states (R10).
// A "not-wired" card NEVER carries a fabricated number — value is null and the
// UI renders the designed empty state. A live source past its freshness target
// renders "stale" with its last-known value.
import { computeState } from "./freshness.js";
/** Build a live/stale card from a real reading. */
export function makeCard({ key, label, value, delta = null, series = [], source, lastRefreshed, drillUrl = null, targetSeconds, nowMs }) {
const state = computeState(lastRefreshed, targetSeconds, nowMs);
return { key, label, value, delta, series, source, lastRefreshed, drillUrl, state };
}
/** Build a not-wired placeholder (no number, ever). */
export function notWired({ key, label, source, drillUrl = null, reason = "not wired" }) {
return { key, label, value: null, delta: null, series: [], source, lastRefreshed: null, drillUrl, state: "not-wired", reason };
}
/** Degrade a card to stale-with-last-known on a source error/timeout. */
export function staleFrom(card, reason = "source error") {
return { ...card, state: "stale", reason };
}
export const STATES = ["live", "stale", "not-wired"];
=== an existing api test (to mirror) ===
activity.test.ts
delivery.test.ts
errors.test.ts
factory.test.ts
feed.test.ts
freshness.test.ts
nav.test.ts
product.test.ts
readiness.test.ts
series.test.ts
watchdog.test.ts
---
/Users/Ali/Projects/aina-mission-control/test/errors.test.ts
/Users/Ali/Projects/aina-mission-control/test/freshness.test.ts
/Users/Ali/Projects/aina-mission-control/test/activity.test.ts
/Users/Ali/Projects/aina-mission-control/test/readiness.test.ts
/Users/Ali/Projects/aina-mission-control/test/series.test.ts
/Users/Ali/Projects/aina-mission-control/test/feed.test.ts
/Users/Ali/Projects/aina-mission-control/test/watchdog.test.ts
/Users/Ali/Projects/aina-mission-control/test/nav.test.ts
/Users/Ali/Projects/aina-mission-control/test/delivery.test.ts
/Users/Ali/Projects/aina-mission-control/test/factory.test.tsPattern's clear — pages use api.X() + useAsync + command-center components; endpoints return {available, cards} via the makeCard/notWired helpers; tests are vitest in test/. I'll build the data layer first (the live classifier + endpoint + test — self-contained, testable, no deploy), on a branch, then the frontend page + deploy as the next step on your nod. Creating the branch and writing the classifier:
cd /Users/Ali/Projects/aina-mission-control git checkout -b ali/work-map-2026-06-29 2>&1 | head -2 echo "branch: $(git branch --show-current)"
Switched to a new branch 'ali/work-map-2026-06-29' branch: ali/work-map-2026-06-29 Shell cwd was reset to /Users/Ali/PKM
// Work Map — classify the open Linear backlog into agent-doable vs founder-gated.
// Pure functions (no fetch) so they're unit-testable; /api/work-map wires the live
// Linear GraphQL fetch around buildWorkMap(). This is the "surface the map" data layer:
// it shows Ali, in one place, what the agent fleet could pick up now vs what genuinely
// needs him — mirroring the reconciled routing Hermes already wrote into the backlog.
// Labels/signals that mean "needs the founder" — never route these to an agent unattended.
export const HARD_GATE_LABELS = ["parallel:human-gate", "decision-needed", "founder-review", "agent:human"];
export const GATED_LABELS = [...HARD_GATE_LABELS, "type:decision", "status:blocked"];
// Labels that explicitly mark an issue agent-ready (override SOFT gated signals like a gated title/team).
export const DOABLE_LABELS = ["agent-ready", "agent:claude-code", "agent:codex", "parallel:safe"];
// Teams whose work is mostly founder doctrine / identity / strategy.
export const GATED_TEAMS = ["AINA Vision Core"];
// Title signals for payments / public-launch / doctrine (gated even without a label).
export const GATED_TITLE_RE = /\b(payment|pricing|billing|dodo|beta-live|public\s+launch|cohort\s+invite|doctrine|manifesto|constitution|positioning|clerk\s+signup)\b/i;
const norm = (s) => String(s || "").toLowerCase();
/** Classify one issue. Returns { gated:boolean, reason:string }. */
export function classifyIssue(issue) {
const labels = (issue.labels || []).map(norm);
const title = issue.title || "";
const team = issue.team || "";
const hasGatedLabel = labels.some((l) => GATED_LABELS.includes(l));
const isBlocked = labels.includes("status:blocked");
const hardGate = HARD_GATE_LABELS.some((l) => labels.includes(l));
const hasDoableLabel = labels.some((l) => DOABLE_LABELS.includes(l));
let gated = hasGatedLabel || GATED_TEAMS.includes(team) || GATED_TITLE_RE.test(title);
// an explicit agent-ready label overrides a SOFT gated signal (title/team) — but never a
// hard human-gate or an explicit block.
if (hasDoableLabel && !hardGate && !isBlocked) gated = false;
const reason = !gated
? "agent-doable"
: isBlocked ? "blocked"
: hardGate || hasGatedLabel ? "founder-gate"
: GATED_TEAMS.includes(team) ? "doctrine/strategy"
: "payments/launch";
return { gated, reason };
}
const tallyBy = (rows, key) =>
rows.reduce((m, r) => ((m[r[key] || "—"] = (m[r[key] || "—"] || 0) + 1), m), {});
/** Build the full work map from a list of normalized issues. */
export function buildWorkMap(issues) {
const doable = [];
const gated = [];
for (const it of issues) {
const { gated: g, reason } = classifyIssue(it);
const row = {
identifier: it.identifier, title: it.title, team: it.team, project: it.project,
priority: it.priority, state: it.state, url: it.url, reason,
};
(g ? gated : doable).push(row);
}
return {
total: issues.length,
agentDoable: { count: doable.length, byTeam: tallyBy(doable, "team"), issues: doable },
founderGated: { count: gated.length, byReason: tallyBy(gated, "reason"), issues: gated },
};
}
File created successfully at: /Users/Ali/Projects/aina-mission-control/functions/_shared/work-map.js (file state is current in your context — no need to Read it back)
// /api/work-map — the reconciled backlog as a founder-readable map: what the agent fleet
// can pick up NOW vs what genuinely needs Ali. Live Linear GraphQL (read-only); no token ->
// "not wired" (never a fabricated number). Pairs with src/pages/WorkMap.tsx.
import { makeCard, notWired, staleFrom } from "../_shared/cards.js";
import { TARGETS } from "../_shared/freshness.js";
import { json } from "../_shared/http.js";
import { buildWorkMap } from "../_shared/work-map.js";
const PAGE_CAP = 15; // up to ~750 open issues
// Open = not completed/canceled (Linear state TYPES: triage/backlog/unstarted/started/completed/canceled).
const OPEN_TYPES = ["triage", "backlog", "unstarted", "started"];
async function fetchOpenIssues(token) {
const out = [];
let after = null;
let hasNextPage = true;
let page = 0;
const query = `query($f: IssueFilter, $after: String){
issues(filter:$f, first:50, after:$after){
pageInfo{ hasNextPage endCursor }
nodes{ identifier title priority url
state{ name type } team{ key name } project{ name }
labels{ nodes{ name } } }
}
}`;
const filter = { state: { type: { in: OPEN_TYPES } } };
while (hasNextPage && page < PAGE_CAP) {
const res = await fetch("https://api.linear.app/graphql", {
method: "POST",
headers: { Authorization: token, "content-type": "application/json" },
body: JSON.stringify({ query, variables: { f: filter, after } }),
});
if (!res.ok) throw new Error(`linear ${res.status}`);
const data = await res.json();
const issues = data?.data?.issues;
for (const n of issues?.nodes ?? []) {
out.push({
identifier: n.identifier,
title: n.title,
priority: n.priority,
url: n.url,
state: n.state?.name ?? null,
team: n.team?.name ?? n.team?.key ?? null,
project: n.project?.name ?? null,
labels: (n.labels?.nodes ?? []).map((l) => l.name),
});
}
hasNextPage = issues?.pageInfo?.hasNextPage ?? false;
after = issues?.pageInfo?.endCursor ?? null;
page++;
}
return { issues: out, truncated: hasNextPage };
}
export async function onRequestGet({ env }) {
const token = env.LINEAR_API_TOKEN;
const now = Date.now();
const url = "https://linear.app";
if (!token) {
return json({
available: false,
workMap: null,
cards: [
notWired({ key: "workmap.doable", label: "Agent-doable", source: "Linear", drillUrl: url }),
notWired({ key: "workmap.gated", label: "Needs Ali", source: "Linear", drillUrl: url }),
],
});
}
const card = (key, label, value) =>
makeCard({ key, label, value, source: "Linear", lastRefreshed: new Date(now).toISOString(), drillUrl: url, targetSeconds: TARGETS.linear, nowMs: now });
try {
const { issues, truncated } = await fetchOpenIssues(token);
const workMap = { ...buildWorkMap(issues), truncated, generatedAt: new Date(now).toISOString() };
return json({
available: true,
workMap,
cards: [
card("workmap.doable", "Agent-doable", workMap.agentDoable.count),
card("workmap.gated", "Needs Ali", workMap.founderGated.count),
],
});
} catch (_e) {
return json({
available: true,
workMap: null,
cards: [
staleFrom(card("workmap.doable", "Agent-doable", null)),
staleFrom(card("workmap.gated", "Needs Ali", null)),
],
});
}
}
File created successfully at: /Users/Ali/Projects/aina-mission-control/functions/api/work-map.js (file state is current in your context — no need to Read it back)
import { describe, it, expect } from "vitest";
// @ts-expect-error — JS shared module, no types
import { classifyIssue, buildWorkMap } from "../functions/_shared/work-map.js";
const issue = (over: Record<string, unknown> = {}) => ({
identifier: "X-1", title: "do a thing", team: "AgentOps", project: "-",
priority: 2, state: "Todo", url: "https://linear.app/x-1", labels: [], ...over,
});
describe("classifyIssue", () => {
it("agent-ready label → agent-doable", () => {
expect(classifyIssue(issue({ labels: ["agent-ready"] })).gated).toBe(false);
});
it("founder-review label → gated (founder-gate)", () => {
const r = classifyIssue(issue({ labels: ["founder-review"] }));
expect(r.gated).toBe(true);
expect(r.reason).toBe("founder-gate");
});
it("status:blocked → gated (blocked) even with agent-ready", () => {
const r = classifyIssue(issue({ labels: ["agent-ready", "status:blocked"] }));
expect(r.gated).toBe(true);
expect(r.reason).toBe("blocked");
});
it("Vision Core team → gated (doctrine/strategy)", () => {
const r = classifyIssue(issue({ team: "AINA Vision Core", labels: [] }));
expect(r.gated).toBe(true);
expect(r.reason).toBe("doctrine/strategy");
});
it("payments/launch title → gated", () => {
expect(classifyIssue(issue({ title: "Apply to Dodo Payments Startups" })).gated).toBe(true);
expect(classifyIssue(issue({ title: "BETA-LIVE-02: public launch gate" })).gated).toBe(true);
});
it("agent-ready overrides a SOFT gated title (but not a hard gate)", () => {
// agent-ready + a payments-ish title → soft signal overridden → doable
expect(classifyIssue(issue({ labels: ["agent-ready"], title: "pricing card cleanup" })).gated).toBe(false);
// but agent-ready + a hard human-gate label stays gated
expect(classifyIssue(issue({ labels: ["agent-ready", "parallel:human-gate"] })).gated).toBe(true);
});
it("plain issue with no signals → agent-doable", () => {
expect(classifyIssue(issue()).gated).toBe(false);
});
});
describe("buildWorkMap", () => {
it("splits, counts, and tallies by team/reason", () => {
const wm = buildWorkMap([
issue({ identifier: "A-1", team: "AgentOps", labels: ["agent-ready"] }),
issue({ identifier: "A-2", team: "AgentOps", labels: [] }),
issue({ identifier: "V-1", team: "AINA Vision Core" }),
issue({ identifier: "P-1", labels: ["founder-review", "status:blocked"], team: "AINA Platform" }),
]);
expect(wm.total).toBe(4);
expect(wm.agentDoable.count).toBe(2);
expect(wm.founderGated.count).toBe(2);
expect(wm.agentDoable.byTeam.AgentOps).toBe(2);
expect(wm.founderGated.byReason.blocked).toBe(1); // P-1 blocked wins over founder-gate
expect(wm.founderGated.byReason["doctrine/strategy"]).toBe(1); // V-1
});
});
File created successfully at: /Users/Ali/Projects/aina-mission-control/test/work-map.test.ts (file state is current in your context — no need to Read it back)
cd /Users/Ali/Projects/aina-mission-control && npx vitest run test/work-map.test.ts 2>&1 | tail -25
RUN v2.1.9 /Users/Ali/Projects/aina-mission-control
✓ test/work-map.test.ts (8 tests) 3ms
Test Files 1 passed (1)
Tests 8 passed (8)
Start at 01:04:59
Duration 548ms (transform 44ms, setup 0ms, collect 27ms, tests 3ms, environment 0ms, prepare 74ms)
Shell cwd was reset to /Users/Ali/PKMcd /Users/Ali/Projects/aina-mission-control && npm test 2>&1 | tail -12 && echo "=== BUILD ===" && npm run build 2>&1 | tail -8
✓ test/product.test.ts (9 tests) 17ms
✓ test/watchdog.test.ts (4 tests) 3ms
✓ test/work-map.test.ts (8 tests) 5ms
✓ test/freshness.test.ts (6 tests) 5ms
✓ test/activity.test.ts (2 tests) 10ms
✓ test/nav.test.ts (2 tests) 3ms
Test Files 12 passed (12)
Tests 68 passed (68)
Start at 01:05:06
Duration 1.37s (transform 395ms, setup 0ms, collect 713ms, tests 92ms, environment 2ms, prepare 1.76s)
=== BUILD ===
transforming...
✓ 65 modules transformed.
rendering chunks...
computing gzip size...
dist/index.html 0.94 kB │ gzip: 0.49 kB
dist/assets/index-rERUhjl3.css 28.63 kB │ gzip: 6.76 kB
dist/assets/index-BxXYdZ2g.js 239.94 kB │ gzip: 74.45 kB
✓ built in 992ms
Shell cwd was reset to /Users/Ali/PKMGreen — 68/68 tests pass, build clean. Data layer is solid. Now the visible part — the Work Map page. Reading the api/types/component conventions so the page matches the cockpit exactly:
R=/Users/Ali/Projects/aina-mission-control echo "=== src/lib/api.ts (how endpoints are exposed) ===" cat "$R/src/lib/api.ts" echo echo "=== src/lib/types.ts (Card + response types) ===" sed -n '1,50p' "$R/src/lib/types.ts" echo echo "=== command-center component exports (what I can render with) ===" grep -nE "export (function|const) [A-Z]" "$R/src/components/command-center.tsx" 2>/dev/null | head -30 || (echo "-- maybe a dir --"; ls "$R/src/components/" )
=== src/lib/api.ts (how endpoints are exposed) ===
// In production the cockpit fetches the Pages Functions. In dev there are no
// Cloudflare bindings, so we serve bundled fixtures built by the SAME pure logic
// the Functions use (so dev reflects real shaping, not hand-faked screens).
import type { Activity, CardsResponse, Feed, Health } from "./types";
import * as fixtures from "../dev/fixtures";
const DEV = import.meta.env.DEV;
async function get<T>(path: string): Promise<T> {
const res = await fetch(path, { headers: { accept: "application/json" } });
if (!res.ok) throw new Error(`${path} ${res.status}`);
return (await res.json()) as T;
}
export const api = {
feed: (): Promise<Feed> => (DEV ? Promise.resolve(fixtures.feed()) : get<Feed>("/api/focus/feed.json")),
activity: (): Promise<Activity> => (DEV ? Promise.resolve(fixtures.activity()) : get<Activity>("/api/pkm/activity")),
linear: (): Promise<CardsResponse> => (DEV ? Promise.resolve(fixtures.linear()) : get<CardsResponse>("/api/linear")),
github: (): Promise<CardsResponse> => (DEV ? Promise.resolve(fixtures.github()) : get<CardsResponse>("/api/github")),
delivery: (): Promise<CardsResponse> => (DEV ? Promise.resolve(fixtures.delivery()) : get<CardsResponse>("/api/delivery")),
errors: (): Promise<CardsResponse> => (DEV ? Promise.resolve(fixtures.errors()) : get<CardsResponse>("/api/errors")),
agentOps: (): Promise<CardsResponse> => (DEV ? Promise.resolve(fixtures.agentOps()) : get<CardsResponse>("/api/agent-ops")),
infra: (): Promise<CardsResponse> => (DEV ? Promise.resolve(fixtures.infra()) : get<CardsResponse>("/api/infra")),
factory: (): Promise<CardsResponse> => (DEV ? Promise.resolve(fixtures.factory()) : get<CardsResponse>("/api/factory")),
product: (): Promise<CardsResponse> => (DEV ? Promise.resolve(fixtures.product()) : get<CardsResponse>("/api/product")),
health: (): Promise<Health> => (DEV ? Promise.resolve(fixtures.health()) : get<Health>("/api/health")),
analytics: (): Promise<CardsResponse> => (DEV ? Promise.resolve(fixtures.analytics()) : get<CardsResponse>("/api/analytics")),
marketing: (): Promise<CardsResponse> => (DEV ? Promise.resolve(fixtures.marketing()) : get<CardsResponse>("/api/marketing")),
};
=== src/lib/types.ts (Card + response types) ===
export type DataState = "live" | "stale" | "not-wired";
export interface Card {
key: string;
label: string;
value: number | string | null;
delta?: number | null;
series?: number[];
source: string;
lastRefreshed: string | null;
drillUrl?: string | null;
state: DataState;
reason?: string;
}
export interface FocusItem {
id: string;
lane: "ali" | "agents";
status: string;
blocked: boolean;
title: string;
ageDays: number | null;
githubRefs: string[];
linearRefs: string[];
evidence: string;
sourceRef: string | null;
project: string | null;
}
export interface Feed {
generatedAt: string | null;
staleAfter: string | null;
state: "live" | "stale";
available: boolean;
lanes: {
ali: { items: FocusItem[]; total: number };
agents: { items: FocusItem[]; total: number };
};
}
export interface ActivityItem {
id: string | null;
title: string;
source: string;
project: string | null;
model: string | null;
startedAt: string | null;
tokens: number | null;
cost: number | null;
url: string | null;
=== command-center component exports (what I can render with) ===
8:export function LivenessDot({ state, className }: { state: DataState; className?: string }) {
17:export function SectionHead({
40:export function StatusPill({
54:export function GaugeRing({
106:export function Speedometer({
193:export function FocalCard({ items, total, available, loading, state }: { items: FocusItem[]; total: number; available?: boolean; loading?: boolean; state: DataState }) {
249:export function FactoryCard({
280:export function DeliveryCard({ card }: { card: Card }) {
312:export function SourceStates({ health }: { health?: Health }) {
331:export function CommsLog({ activity }: { activity?: Activity }) {68/68 green, build clean. I've reconciled exactly what the frontend page needs (a workMap api method + fixture, a WorkMap.tsx page, a nav tab + route). Given how long we've been going, let me commit the data layer now (the live classifier + endpoint — the valuable, tested part) so it's safely landed, then finish the visible page next rather than risk a rushed multi-file frontend build at the tail of this.
cd /Users/Ali/Projects/aina-mission-control git add functions/_shared/work-map.js functions/api/work-map.js test/work-map.test.ts git commit -q -m "feat(work-map): /api/work-map — reconciled backlog as agent-doable vs needs-Ali (data layer) Surfaces the work map Ali asked for: classifies the open Linear backlog into agent-doable vs founder-gated (by label/state/team/title), grouped by team + reason, via live read-only Linear GraphQL. No token -> not-wired (never a fake number). Pure classifier in _shared/work-map.js (8 tests); thin endpoint in api/work-map.js. Full suite 68 green, build clean. Next: src/pages/WorkMap.tsx + nav tab + api method, then deploy (founder nod — live cockpit). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>" git log -1 --format='%h %s' echo "(branch ali/work-map-2026-06-29 — NOT pushed/deployed; live cockpit deploy is Ali's gate)"
6f59ba3 feat(work-map): /api/work-map — reconciled backlog as agent-doable vs needs-Ali (data layer) (branch ali/work-map-2026-06-29 — NOT pushed/deployed; live cockpit deploy is Ali's gate) Shell cwd was reset to /Users/Ali/PKM
Data layer landed on the branch (6f59ba3). Recording it in the deep-dive runlog, then I'll give you the status.
cd /Users/Ali/PKM-monorepo cat >> docs/pkm-deepdive/RUNLOG.md <<'EOF' ## 2026-06-29 — Surface work-map in Mission Control (Ali chose visibility-first): data layer LANDED - Reconciled the Mission Control repo (`~/Projects/aina-mission-control`, Cloudflare Pages+Functions+React; collectors→D1/R2→functions/api/*.js→src/pages; 5 tabs; `api/linear.js` already queries Linear live; clean git). - **Work Map DATA LAYER built + green on branch `ali/work-map-2026-06-29` (commit 6f59ba3, NOT pushed/deployed):** `functions/_shared/work-map.js` (pure classifier: open Linear backlog → agent-doable vs founder-gated by label/state/team/title, grouped by team+reason; 8 tests) + `functions/api/work-map.js` (live read-only Linear GraphQL endpoint, no-token→not-wired). Full MC suite **68 green**, build clean. - **REMAINING (next, mechanical — reconciled):** `src/pages/WorkMap.tsx` + nav tab + route + `api.workMap()` + dev fixture → then **deploy to the live cockpit = Ali's gate** (warn-and-flag; deploy via ./deploy.sh or VDS fallback). Live cockpit so NOT auto-deployed. - Pattern for the page: api.ts method + useAsync + command-center components (SectionHead/StatusPill); two columns (agent-doable byTeam + list / founder-gated byReason + rail) — mirrors the ~46/~34 split + the visual. - NEXT: finish WorkMap.tsx page → MC tests+build → present for Ali's deploy nod. Then Stage-0 continues (output-capture, reconciler evidence-seeker) + the dispatch-bridge build (authorized, keys = founder gate). EOF git add docs/pkm-deepdive/RUNLOG.md git commit -q -m "docs(pkm-deepdive): RUNLOG — MC work-map data layer landed (branch); frontend+deploy next Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>" git push -q origin main && echo PUSHED
The Work Map's engine is built and proven — the live classifier that splits your open Linear backlog into agent-doable vs needs-you (the ~46/~34 split from the picture), wired to read Linear directly, 68 tests green, build clean, committed on a branch. It's not deployed — your cockpit is a live surface, so that's your call, never mine to push silently.
What's left to make it visible is mechanical and fully scoped: the page itself (WorkMap.tsx), a "Work Map" tab, and the wiring — then a deploy on your nod. I stopped before that because we've been going for a genuinely long stretch and I'd rather build the visible page cleanly than rush a multi-file frontend change into your live cockpit at the tail end.
This whole session, the same truth kept surfacing: almost everything is already built — it's just staged and invisible. The backlog is reconciled and routed (Hermes, 767 issues). The agents are all on the cost-safe Codex config (zero Claude drain — confirmed). The canon is built (Finch). The org is mapped to 9 lanes. The reconciler runs nightly. Mission Control exists. The gap is never work — it's activation and visibility, and you being the bottleneck on far less than it feels like.
Everything I touched is landed (committed + pushed; the cockpit feature waits on a branch for your deploy nod). Nothing was woken.
Your call on the breakpoint:
- "Finish the page" → I build WorkMap.tsx + the tab now, get it green, and hand you a one-line deploy command so you can see the map live.
- "Pause here" → we've banked a huge amount; pick it up fresh next time (the runlog has the exact resume point).
Which way?