cd /Users/Ali/PKM/scratch/pkm-deepdive
cat > /tmp/reconcile-frodo.sh <<'A'
CID=7d58fc13-7c9a-4c87-b9f8-a7cfae8564a9
echo "===== 1. Frodo / Gimli / release-manager agents in Paperclip ====="
paperclipai agent list -C "$CID" --json 2>/dev/null | python3 -c "
import json,sys
d=json.load(sys.stdin); a=d if isinstance(d,list) else d.get('agents',[])
for x in a:
  nm=(x.get('name') or '').lower(); role=(x.get('role') or x.get('title') or '')
  if any(k in nm for k in ('frodo','gimli','sam','release','marshal','pippin','merry','aragorn','legolas','gandalf')) or 'release' in role.lower():
    print('  ',x.get('name'),'|',role,'|',x.get('status'),'|',x.get('id','')[:8])
print('  --- total agents:',len(a))
" 2>/dev/null
echo
echo "===== 2. search Frodo/release across ops + docs ====="
grep -rliE "frodo|release manager|release-manager|promote.*main|dev.*main|batch promot" /srv/aina/ops /srv/repos/aina-academy/docs 2>/dev/null | head -10 | sed 's/^/  /'
echo
echo "===== 3. dev branch / integration branch exist? ====="
cd /srv/repos/aina-academy 2>/dev/null && git branch -a 2>/dev/null | grep -iE "dev|integration|staging|release" | head | sed 's/^/  /'
echo "  remote branches:"; git ls-remote --heads origin 2>/dev/null | awk '{print $2}' | grep -iE "dev|integration|staging|release" | head | sed 's/^/  /'
echo
echo "===== 4. teams: dev vs marketing ====="
paperclipai team list -C "$CID" --json 2>/dev/null | python3 -c "import json,sys;d=json.load(sys.stdin);t=d if isinstance(d,list) else d.get('teams',[]);[print('  ',x.get('name'),x.get('id','')[:8]) for x in t]" 2>/dev/null | head -15 || echo "  (no team list cmd)"
echo
echo "===== 5. CURRENT watchers/bots/crons to remove ====="
crontab -l 2>/dev/null | grep -vE '^#|^$' | sed 's/^/  cron: /'
echo "  --- github workflows (bots):"; ls /srv/repos/aina-academy/.github/workflows/*.yml 2>/dev/null | xargs -n1 basename | sed 's/^/    /'
A
B64=$(base64 < /tmp/reconcile-frodo.sh | tr -d '\n')
./vds-exec.sh run "echo $B64 | base64 -d | bash" 2>&1

cd /Users/Ali/PKM/scratch/pkm-deepdive
cat > /tmp/teardown.sh <<'TD'
set -uo pipefail
SELF=$$
echo "===== 1. Kill running merge-train / PR-merge lanes ====="
for p in $(pgrep -f "lane-merge-train\|codex-merge-train" 2>/dev/null); do [ "$p" != "$SELF" ] && kill "$p" 2>/dev/null && echo "  killed $p"; done
echo "  done"
echo
echo "===== 2. Pause GitHub/watcher/producer crons (comment out; keep janitor+poller) ====="
crontab -l 2>/dev/null > /tmp/ct.bak
python3 - <<'PY'
lines=open('/tmp/ct.bak').read().splitlines(); out=[]
for l in lines:
    if any(s in l for s in ('release-marshal.sh','watchdog.sh','coo-ops-loop.sh')) and not l.lstrip().startswith('#'):
        out.append('# PAUSED 2026-06-30 (Ali: remove GitHub/watchers/bots; Frodo does releases): '+l)
    else:
        out.append(l)
open('/tmp/ct.new','w').write('\n'.join(out)+'\n')
PY
crontab /tmp/ct.new
echo "  active crons now:"; crontab -l 2>/dev/null | grep -vE '^#|^$' | sed 's/^/    /'
echo
echo "===== 3. Disable GitHub bot workflows ====="
cd /srv/repos/aina-academy 2>/dev/null || exit 1
for wf in pr-bot-review-watcher.yml rollout-health-monitor.yml; do
  gh workflow disable "$wf" 2>&1 | head -1 | sed "s/^/  $wf: /" || echo "  $wf: (already off or n/a)"
done
echo
echo "===== 4. hermes gateway: systemd auto-restart (so no watchdog needed for it) ====="
systemctl --user show hermes-gateway-aina-chief-of-staff.service -p Restart 2>/dev/null | sed 's/^/  /'
echo
echo "===== STATE ====="
echo "  merge-train procs: $(pgrep -f 'lane-merge-train' | grep -v $SELF | wc -l)"
echo "  release-marshal/watchdog/coo crons: PAUSED"
TD
B64=$(base64 < /tmp/teardown.sh | tr -d '\n')
./vds-exec.sh runl "echo $B64 | base64 -d | bash" 2>&1

cd /Users/Ali/PKM/scratch/pkm-deepdive
cat > /tmp/frodo-design.sh <<'A'
CID=7d58fc13-7c9a-4c87-b9f8-a7cfae8564a9
echo "===== Frodo agent config / instructions ====="
paperclipai agent get 88b49386 -C "$CID" --json 2>/dev/null | python3 -c "import json,sys;d=json.load(sys.stdin);print('  name:',d.get('name'));print('  role:',d.get('role'));print('  status:',d.get('status'));p=(d.get('systemPrompt') or d.get('prompt') or d.get('instructions') or '');print('  prompt/instructions (first 800):');print('   '+p[:800].replace(chr(10),'\n   '))" 2>/dev/null || paperclipai agent get 88b49386 -C "$CID" 2>/dev/null | head -30 | sed 's/^/  /'
echo
echo "===== release-flow test runs (docs/logs) ====="
ls -t /srv/aina/ops/*release* /srv/aina/ops/*frodo* /srv/aina/ops/*promot* 2>/dev/null | head | sed 's/^/  /'
grep -rliE "frodo.*release|release manager|dev.*to.*main|promote.*milestone|preview url|wrangler.*preview|pages deploy" /srv/repos/aina-academy/docs /srv/aina/ops 2>/dev/null | head -8 | sed 's/^/  /'
echo
echo "===== preview deploy mechanism (wrangler / cloudflare preview) ====="
cd /srv/repos/aina-academy 2>/dev/null && grep -rlE "wrangler|pages deploy|preview" package.json .github/workflows/*.yml scripts/ 2>/dev/null | head | sed 's/^/  /'
grep -hE "\"deploy|preview|wrangler" package.json 2>/dev/null | head -8 | sed 's/^/  /'
echo
echo "===== verify bots disabled ====="
gh api repos/:owner/:repo/actions/workflows --jq '.workflows[]|select(.name|test("bot|watcher|rollout";"i"))|"  \(.name): \(.state)"' 2>/dev/null | head
A
B64=$(base64 < /tmp/frodo-design.sh | tr -d '\n')
./vds-exec.sh run "echo $B64 | base64 -d | bash" 2>&1

cd /Users/Ali/PKM/scratch/pkm-deepdive
cat > /tmp/readspec.sh <<'A'
F=/srv/repos/aina-academy/docs/ops/aina-agent-factory-spec-2026-06-29.html
python3 - "$F" <<'PY'
import sys,re,html
t=open(sys.argv[1],encoding='utf-8',errors='ignore').read()
t=re.sub(r'<script.*?</script>','',t,flags=re.S); t=re.sub(r'<style.*?</style>','',t,flags=re.S)
txt=html.unescape(re.sub(r'<[^>]+>',' ',t))
txt=re.sub(r'[ \t]+',' ',txt)
lines=[l.strip() for l in txt.splitlines() if l.strip()]
full=' '.join(lines)
# find sentences mentioning key concepts
for kw in ['Frodo','release','promot','dev branch','main','milestone','epic','preview','marketing','deploy','internal']:
    print(f"=== '{kw}' ===")
    for m in re.finditer(r'[^.]*\b'+re.escape(kw)+r'[^.]*\.', full, re.I):
        s=m.group(0).strip()
        if 25<len(s)<320: print('  '+s[:300])
    print()
PY
A
B64=$(base64 < /tmp/readspec.sh | tr -d '\n')
./vds-exec.sh run "echo $B64 | base64 -d | bash" 2>&1 | head -70

cd /Users/Ali/PKM/scratch/pkm-deepdive
cat > /tmp/recon2.sh <<'A'
CID=7d58fc13-7c9a-4c87-b9f8-a7cfae8564a9
echo "===== PAUSED STATE HOLDING? ====="
echo "  active crons: $(crontab -l 2>/dev/null | grep -vE '^#|^$' | grep -cE 'watchdog|release-marshal|coo-ops')  (want 0 of watchdog/marshal/coo)"
echo "  codex lanes running: $(pgrep -f 'codex exec'|wc -l)"
echo "  new PRs since teardown (21:00Z+): $(cd /srv/repos/aina-academy 2>/dev/null && gh pr list --state open --json createdAt --jq '[.[]|select(.createdAt>"2026-07-01T01:00:00Z")]|length' 2>/dev/null)"
echo
echo "===== DEV vs MARKETING structure: Paperclip goals ====="
paperclipai goal list -C "$CID" --json 2>/dev/null | python3 -c "import json,sys;d=json.load(sys.stdin);g=d if isinstance(d,list) else d.get('goals',[]);[print('  ',(x.get('name') or x.get('title') or '')[:55],'|',x.get('id','')[:8]) for x in g]" 2>/dev/null | head -25 || echo "  (no goal list)"
echo
echo "===== Frodo / release test runs: git history ====="
cd /srv/repos/aina-academy 2>/dev/null && git log --all --oneline --format='  %ai %s' 2>/dev/null | grep -iE "frodo|release manager|promote.*main|dev.*main|milestone.*promot|batch.*promot" | head -8
echo "  branches mentioning release/promote/frodo:"; git branch -a 2>/dev/null | grep -iE "frodo|release|promot|dev-to-main" | head
echo
echo "===== docs mentioning the release/promotion flow ====="
grep -rliE "frodo|release manager|promote.*to.*main|dev.*branch.*main|batch promotion|milestone.*deploy" /srv/repos/aina-academy/docs /srv/aina/ops /srv/repos/.fusion 2>/dev/null | head -10 | sed 's/^/  /'
A
B64=$(base64 < /tmp/recon2.sh | tr -d '\n')
./vds-exec.sh run "echo $B64 | base64 -d | bash" 2>&1

cat >> /Users/Ali/PKM/scratch/pkm-deepdive/EXECUTION-STATE-2026-06-30.md <<'EOF'

## MAJOR DIRECTION CHANGE (Ali, ~evening): GITHUB-PER-TASK WAS DRIFT — go internal/hybrid
Ali's correction: system was DESIGNED as hybrid — heavy work fast+contained locally, GitHub main only after MILESTONES/EPICS (main=deploy source), NOT per-task. Per-task PR/CI/review/merge drifted in over ~10 days (proof-rails+canon-guards+pr-bot-watcher #123). I spent today DEEPENING the wrong model (Mergify/merge-train/release-marshal) instead of questioning it.
DIRECTIVES: (1) release mgmt = FRODO agent (88b49386 devops, idle) not my cron scripts — "factory does it itself"; Gimli(2fe6579c qa)=QA. (2) promote after milestones/epics not every task. (3) DEV TEAM ONLY (=Journey/Curriculum/Data/QA goals), NOT marketing. (4) remove GitHub part NOW. (5) keep internal; share Cloudflare PREVIEW URLs (wrangler) until work done. (6) NO more watchers and bots.
DONE (teardown): killed merge-train; PAUSED crons release-marshal+watchdog+coo-ops; DISABLED bots pr-bot-review-watcher+rollout-health-monitor (disabled_manually); hermes now systemd Restart=always (no watchdog needed). Production PAUSED. 12 in-flight lanes from 18:15 cycle finishing on OLD model (will push ~final PRs).
MODEL TO BUILD (Frodo internal): dev-team agents build in local worktrees→commit local `dev` branch (no push/PR/CI per-task); Frodo builds dev + wrangler Cloudflare PREVIEW → shares URL w/ Ali; at milestone Frodo integrates + Gimli QA verifies → on Ali's go promote dev→GitHub main→prod deploy. NO cron scripts — Frodo the agent runs it.
BLOCKED ON ALI: point me to the release-flow TEST RUNS (he said "ran a couple") to restore his exact design. Dev-team scope RESOLVED (Journey/Curriculum/Data/QA).
DO NOT rebuild on GitHub-per-task. Retire (don't just pause) release-marshal/merge-train/Mergify-queue once Frodo flow proven.
EOF
echo "logged"

cd /Users/Ali/PKM/scratch/pkm-deepdive
cat > /tmp/pc-dump.sh <<'A'
CID=7d58fc13-7c9a-4c87-b9f8-a7cfae8564a9
echo "===== issue list schema (one issue, all fields) ====="
paperclipai issue list -C "$CID" --json 2>/dev/null | python3 -c "import json,sys;d=json.load(sys.stdin);i=d if isinstance(d,list) else d.get('issues',[]);print('  total issues:',len(i));print('  fields:',sorted((i[0].keys())) if i else 'none')" 2>/dev/null
echo
echo "===== dump full issue list to file for analysis ====="
paperclipai issue list -C "$CID" --json 2>/dev/null > /srv/aina/ops/pc-issues-full.json
echo "  dumped: $(wc -c < /srv/aina/ops/pc-issues-full.json) bytes"
echo
echo "===== one issue's FULL detail (to see verification/comment/history structure) ====="
FIRST=$(python3 -c "import json;d=json.load(open('/srv/aina/ops/pc-issues-full.json'));i=d if isinstance(d,list) else d.get('issues',[]);print([x for x in i if x.get('status')=='done'][0]['id'])" 2>/dev/null)
echo "  sample done issue: $FIRST"
paperclipai issue get "$FIRST" -C "$CID" --json 2>/dev/null | python3 -c "import json,sys;d=json.load(sys.stdin);print('  keys:',sorted(d.keys()));[print('   ',k,'=',str(d[k])[:120]) for k in ('title','status','assigneeAgentId','verifierAgentId','builderAgentId','createdAt','completedAt','goalId') if k in d]" 2>/dev/null
echo
echo "===== agents id->name map ====="
paperclipai agent list -C "$CID" --json 2>/dev/null | python3 -c "import json,sys;d=json.load(sys.stdin);a=d if isinstance(d,list) else d.get('agents',[]);[print('  ',x.get('id','')[:8],x.get('name'),'|',x.get('role')) for x in a if any(k in (x.get('name') or '').lower() for k in ('gimli','frodo','gandalf','legolas','jessica','coo','harvey','jared'))]" 2>/dev/null
A
B64=$(base64 < /tmp/pc-dump.sh | tr -d '\n')
./vds-exec.sh run "echo $B64 | base64 -d | bash" 2>&1

cd /Users/Ali/PKM/scratch/pkm-deepdive
./vds-exec.sh run "cat /srv/aina/ops/pc-issues-full.json" > /tmp/pc-issues.json 2>/dev/null
echo "pulled $(wc -c < /tmp/pc-issues.json) bytes"
python3 - <<'PY'
import json,collections,datetime
d=json.load(open('/tmp/pc-issues.json'))
issues=d if isinstance(d,list) else d.get('issues',[])
AG={'a873590c':'Jared','85254289':'Harvey','60554106':'COO','6454b8e0':'Jessica','af994de7':'Legolas','a40a26d9':'Gandalf','2fe6579c':'Gimli','88b49386':'Frodo'}
def an(x): 
    x=(x or '')[:8]; return AG.get(x,x[:8] if x else '-')
print("TOTAL issues:",len(issues))
print("\n=== BY STATUS ===")
for s,c in collections.Counter(i.get('status') for i in issues).most_common(): print(f"  {s}: {c}")
print("\n=== CREATED per day (factory timeline) ===")
byday=collections.Counter((i.get('createdAt') or '')[:10] for i in issues)
for day in sorted(byday): 
    if day: print(f"  {day}: {byday[day]} created")
print("\n=== COMPLETED per day ===")
compday=collections.Counter((i.get('completedAt') or '')[:10] for i in issues if i.get('completedAt'))
for day in sorted(compday):
    if day: print(f"  {day}: {compday[day]} completed")
print("\n=== assigneeAgentId (who OWNS tasks) top ===")
for a,c in collections.Counter(an(i.get('assigneeAgentId')) for i in issues).most_common(12): print(f"  {a}: {c}")
print("\n=== executionAgentNameKey (who EXECUTED) top ===")
for a,c in collections.Counter((i.get('executionAgentNameKey') or '-') for i in issues).most_common(12): print(f"  {a}: {c}")
print("\n=== createdByAgentId (who CREATED tasks) ===")
for a,c in collections.Counter(an(i.get('createdByAgentId')) for i in issues).most_common(8): print(f"  {a}: {c}")
print("\n=== workMode / executionPolicy ===")
for a,c in collections.Counter(i.get('workMode') for i in issues).most_common(): print(f"  workMode {a}: {c}")
PY

cd /Users/Ali/PKM/scratch/pkm-deepdive
python3 - <<'PY'
import json,collections
d=json.load(open('/tmp/pc-issues.json'))
issues=d if isinstance(d,list) else d.get('issues',[])
AG={'a873590c':'Jared','85254289':'Harvey','60554106':'COO','6454b8e0':'Jessica','af994de7':'Legolas','a40a26d9':'Gandalf','2fe6579c':'Gimli','88b49386':'Frodo'}
def an(x): x=(x or '')[:8]; return AG.get(x,x if x else '-')
# in_review vs done: created date distribution (is in_review the 06-29 flood?)
print("=== in_review issues by created-day (are they the 06-29 flood?) ===")
ir=[i for i in issues if i.get('status')=='in_review']
for day,c in sorted(collections.Counter((i.get('createdAt') or '')[:10] for i in ir).items()): print(f"  {day}: {c}")
print(f"  in_review assignees:", dict(collections.Counter(an(i.get('assigneeAgentId')) for i in ir)))
print("\n=== done issues by created-day (clean early period?) ===")
dn=[i for i in issues if i.get('status')=='done']
for day,c in sorted(collections.Counter((i.get('createdAt') or '')[:10] for i in dn).items()): print(f"  {day}: {c}")
print("\n=== Gimli's 9 issues: what are they ===")
for i in issues:
    if an(i.get('assigneeAgentId'))=='Gimli': print(f"  [{i.get('status')}] {(i.get('title') or '')[:70]}")
print("\n=== labels across all issues (verification signal?) ===")
labs=collections.Counter()
for i in issues:
    for l in (i.get('labels') or []): labs[l.get('name') if isinstance(l,dict) else l]+=1
for l,c in labs.most_common(15): print(f"  {l}: {c}")
print("\n=== early issues (06-19..06-28) vs flood (06-29+): completion ===")
early=[i for i in issues if (i.get('createdAt') or '')[:10] <= '2026-06-28']
flood=[i for i in issues if (i.get('createdAt') or '')[:10] >= '2026-06-29']
print(f"  EARLY (<=06-28): {len(early)} total, {sum(1 for i in early if i.get('status')=='done')} done, {sum(1 for i in early if i.get('status')=='in_review')} in_review")
print(f"  FLOOD (>=06-29): {len(flood)} total, {sum(1 for i in flood if i.get('status')=='done')} done, {sum(1 for i in flood if i.get('status')=='in_review')} in_review, {sum(1 for i in flood if i.get('status')=='todo')} todo")
PY

cd /Users/Ali/PKM/scratch/pkm-deepdive
python3 - <<'PY'
import json,collections
d=json.load(open('/tmp/pc-issues.json'))
issues=d if isinstance(d,list) else d.get('issues',[])
def labs(i): return [ (l.get('name') if isinstance(l,dict) else l) for l in (i.get('labels') or []) ]
print("=== 'release' labeled issues (the test runs?) ===")
for i in issues:
    if 'release' in labs(i): print(f"  [{i.get('status')}] {(i.get('createdAt') or '')[:10]} {(i.get('title') or '')[:75]}")
print("\n=== milestone (M0-M8) labels: issues per milestone + done ===")
ms=collections.defaultdict(lambda:[0,0])
for i in issues:
    for l in labs(i):
        if l and l.startswith('M') and len(l)<=3 and l[1:].isdigit():
            ms[l][0]+=1
            if i.get('status')=='done': ms[l][1]+=1
for m in sorted(ms): print(f"  {m}: {ms[m][0]} issues, {ms[m][1]} done")
print("\n=== EARLY period titles (06-19..06-28) — what the clean factory built ===")
early=sorted([i for i in issues if (i.get('createdAt') or '')[:10]<='2026-06-28'], key=lambda x:x.get('createdAt') or '')
for i in early[:18]: print(f"  {(i.get('createdAt') or '')[:10]} [{i.get('status')}] {(i.get('title') or '')[:65]}")
print("\n=== sample EARLY done issue: description (shows verify flow?) ===")
samp=[i for i in early if i.get('status')=='done' and 'verif' in (i.get('description') or i.get('title') or '').lower()]
if not samp: samp=[i for i in early if i.get('status')=='done']
if samp:
    i=samp[0]; print(f"  title: {i.get('title')}"); print(f"  desc: {(i.get('description') or '')[:400]}")
PY

cd /Users/Ali/PKM/scratch/pkm-deepdive
cat > /tmp/verifier-recon.sh <<'A'
CID=7d58fc13-7c9a-4c87-b9f8-a7cfae8564a9
echo "===== paperclipai agent subcommands (create/clone?) ====="
paperclipai agent --help 2>&1 | grep -iE "create|clone|add|new|update|spawn" | head | sed 's/^/  /'
echo
echo "===== paperclipai issue update options (how to mark done/verified) ====="
paperclipai issue update --help 2>&1 | grep -iE "status|assign|state|done|comment" | head | sed 's/^/  /'
echo
echo "===== Gimli full config (to clone role/prompt) ====="
paperclipai agent get 2fe6579c -C "$CID" 2>&1 | head -40 | sed 's/^/  /'
echo
echo "===== sample in_review issue: what verifying it needs ====="
python3 -c "
import json;d=json.load(open('/tmp/pc-issues.json'));i=d if isinstance(d,list) else d.get('issues',[])
ir=[x for x in i if x.get('status')=='in_review']
x=ir[0]
print('  id:',x.get('id')); print('  identifier:',x.get('identifier')); print('  title:',x.get('title'))
print('  goalId:',(x.get('goalId') or '')[:8]); print('  executionState:',x.get('executionState'))
print('  desc:',(x.get('description') or '')[:300])
"
A
B64=$(base64 < /tmp/verifier-recon.sh | tr -d '\n')
./vds-exec.sh run "echo $B64 | base64 -d | bash" 2>&1

cd /Users/Ali/PKM/scratch/pkm-deepdive
echo "===== LOCAL: do the 70 in_review tasks have BUILT work to verify? ====="
python3 - <<'PY'
import json,collections
d=json.load(open('/tmp/pc-issues.json'))
i=d if isinstance(d,list) else d.get('issues',[])
ir=[x for x in i if x.get('status')=='in_review']
print("  in_review total:",len(ir))
print("  executionState:",dict(collections.Counter(x.get('executionState') for x in ir)))
print("  have executionRunId (were executed/built):",sum(1 for x in ir if x.get('executionRunId')))
print("  have startedAt:",sum(1 for x in ir if x.get('startedAt')))
print("  by goal-prefix (dev teams):")
# map goalId->name via a quick guess from title prefix
pref=collections.Counter()
for x in ir:
    t=(x.get('title') or '')
    p=t.split(':')[0].split('-')[0].strip()[:14] if t else '?'
    pref[p]+=1
for p,c in pref.most_common(10): print(f"    {p}: {c}")
print("\n  sample 3 in_review (identifier | title | executionState):")
for x in ir[:3]:
    print(f"    {x.get('identifier')} | {(x.get('title') or '')[:55]} | exec={x.get('executionState')} run={bool(x.get('executionRunId'))}")
PY
echo
echo "===== VDS: agent create syntax + Gimli config ====="
cat > /tmp/vr2.sh <<'A'
echo "-- agent create --help:"; paperclipai agent create --help 2>&1 | head -20 | sed 's/^/  /'
echo "-- agent get syntax:"; paperclipai agent get --help 2>&1 | head -8 | sed 's/^/  /'
A
B64=$(base64 < /tmp/vr2.sh | tr -d '\n')
./vds-exec.sh run "echo $B64 | base64 -d | bash" 2>&1

cd /Users/Ali/PKM/scratch/pkm-deepdive
# get in_review identifiers locally
python3 -c "
import json;d=json.load(open('/tmp/pc-issues.json'));i=d if isinstance(d,list) else d.get('issues',[])
ir=[x.get('identifier') for x in i if x.get('status')=='in_review' and x.get('identifier')]
print(' '.join(ir))
" > /tmp/ir-ids.txt
echo "in_review identifiers: $(wc -w < /tmp/ir-ids.txt)"
IDS=$(cat /tmp/ir-ids.txt)
cat > /tmp/checkbuilt.sh <<A
cd /srv/repos/aina-academy 2>/dev/null || exit
git fetch -q origin 2>/dev/null
echo "-- for each in_review identifier: is there a branch or PR (=built)?"
built=0; unbuilt=0
for id in $IDS; do
  low=\$(echo "\$id" | tr 'A-Z' 'a-z')
  br=\$(git ls-remote --heads origin 2>/dev/null | grep -iE "\$low|\${id}" | head -1)
  pr=\$(gh pr list --state all --search "\$id" --json number --jq '.[0].number' 2>/dev/null)
  if [ -n "\$br" ] || [ -n "\$pr" ]; then built=\$((built+1)); else unbuilt=\$((unbuilt+1)); fi
done
echo "  BUILT (has branch/PR): \$built"
echo "  UNBUILT (no branch/PR): \$unbuilt"
A
B64=$(base64 < /tmp/checkbuilt.sh | tr -d '\n')
./vds-exec.sh run "echo $B64 | base64 -d | bash" 2>&1

cd /Users/Ali/PKM/scratch/pkm-deepdive
cat > /tmp/gimli-cfg.sh <<'A'
CID=7d58fc13-7c9a-4c87-b9f8-a7cfae8564a9
echo "===== Gimli full config (template for clones) ====="
paperclipai agent get 2fe6579c --company-id "$CID" --json 2>/dev/null | python3 -c "
import json,sys
d=json.load(sys.stdin)
print('  keys:',sorted(d.keys()))
for k in ('name','role','status','model','adapterConfigKey','instructionsPath','instructionsBundle','permissions','executionPolicy','workMode'):
    if k in d: print(f'  {k}: {str(d[k])[:200]}')
" 2>&1 | head -30
echo
echo "===== how issue<->branch links (identifier in branch/PR)? sample ====="
cd /srv/repos/aina-academy 2>/dev/null && gh pr list --state open --json number,title,headRefName --jq '.[0:3][]|"  #\(.number) \(.headRefName) — \(.title[0:40])"' 2>/dev/null
A
B64=$(base64 < /tmp/gimli-cfg.sh | tr -d '\n')
./vds-exec.sh run "echo $B64 | base64 -d | bash" 2>&1

cd /Users/Ali/PKM/scratch/pkm-deepdive
# Build identifier->internal-id map + 3 batches locally
python3 - <<'PY'
import json
d=json.load(open('/tmp/pc-issues.json'))
i=d if isinstance(d,list) else d.get('issues',[])
ir=[(x.get('identifier'),x.get('id')) for x in i if x.get('status')=='in_review' and x.get('identifier') and x.get('id')]
m={ident:iid for ident,iid in ir}
json.dump(m,open('mergify/ir-map.json','w'))  # reuse dir
ids=[ident for ident,_ in ir]
# 3 balanced batches
batches=[ids[j::3] for j in range(3)]
for n,b in enumerate(batches,1):
    open(f'mergify/ir-batch{n}.txt','w').write(' '.join(b))
    print(f"batch{n}: {len(b)} ids")
print("map entries:",len(m))
PY
echo "---prompt---"
cat > mergify/verifier-prompt.txt <<'PROMPT'
You are a parallel adversarial VERIFIER (a "Gimli") for the AINA agent factory, clearing the in_review backlog INTERNALLY (no GitHub merges). Company CID=7d58fc13-7c9a-4c87-b9f8-a7cfae8564a9. The identifier->internal-id map is /srv/aina/ops/ir-map.json. Your assigned batch of identifiers is in the env var BATCH (space-separated).

For EACH identifier in BATCH:
1. internal id = ir-map.json[identifier]. Read the Paperclip issue: `paperclipai issue get <internal-id> --company-id $CID` — understand its acceptance criteria / definition of done.
2. Find its built work in /srv/repos/aina-academy: `gh pr list --state all --search "<identifier>" --json number,state,headRefName` then `gh pr diff <n>` (or inspect the branch). 
3. ADVERSARIALLY review the diff against the acceptance criteria. Actively try to find why it should NOT pass: stubs, TODOs, missing tests, doesn't meet the spec, unsafe, incomplete. You are a skeptic — this is adversarial verification, NOT rubber-stamping.
4. Verdict via paperclipai (use internal-id + --company-id $CID):
   - PASS (genuinely satisfies acceptance): `paperclipai issue update <internal-id> --company-id $CID --status done --comment "verified (parallel-Gimli): <1-line concrete evidence>"`
   - FAIL (stub/incomplete/unsafe/doesn't meet acceptance): `... --status todo --comment "rejected (parallel-Gimli): <specific gap>"`
   - NO built work (no PR/branch): `... --status todo --comment "no built work found — needs building"`
5. Honest + specific. Do NOT mark done unless it genuinely meets acceptance.

DO NOT: merge PRs, push, edit code (you VERIFY only). NEVER gpt-5.3-codex-spark. When batch done, summarize: N done, M rejected, K no-work.
PROMPT
wc -l mergify/verifier-prompt.txt; echo "batches + map + prompt ready"

cd /Users/Ali/PKM/scratch/pkm-deepdive
MAP_B64=$(base64 < mergify/ir-map.json | tr -d '\n')
PROMPT_B64=$(base64 < mergify/verifier-prompt.txt | tr -d '\n')
B1=$(cat mergify/ir-batch1.txt); B2=$(cat mergify/ir-batch2.txt); B3=$(cat mergify/ir-batch3.txt)
cat > /tmp/launch-verifiers.sh <<LAUNCH
set -uo pipefail
export PATH="/home/ali/.npm-global/bin:/usr/local/bin:/usr/bin:/bin:\$PATH"
CID=7d58fc13-7c9a-4c87-b9f8-a7cfae8564a9
mkdir -p /srv/aina/ops/verifiers
echo "$MAP_B64" | base64 -d > /srv/aina/ops/ir-map.json
echo "$PROMPT_B64" | base64 -d > /srv/aina/ops/verifiers/verifier-prompt.txt
CODEX=\$(command -v codex || echo /home/ali/.npm-global/bin/codex)
launch(){ # \$1=lane-num \$2=batch-ids
  setsid nohup env BATCH="\$2" CID="\$CID" bash -c "cat /srv/aina/ops/verifiers/verifier-prompt.txt | \$CODEX exec --json --dangerously-bypass-approvals-and-sandbox --model gpt-5.5 -c model_reasoning_effort=high --skip-git-repo-check - > /srv/aina/ops/verifiers/gimli-\$1.log 2>&1" >/dev/null 2>&1 &
  echo "  launched Gimli-\$1 (\$(echo \$2|wc -w) tasks)"
}
launch 1 "$B1"
launch 2 "$B2"
launch 3 "$B3"
sleep 3
echo "-- verifier lanes running:"; pgrep -fa "verifier-prompt" | grep -v pgrep | wc -l | sed 's/^/  /'
LAUNCH
B64=$(base64 < /tmp/launch-verifiers.sh | tr -d '\n')
./vds-exec.sh runl "echo $B64 | base64 -d | bash" 2>&1

cat >> /Users/Ali/PKM/scratch/pkm-deepdive/EXECUTION-STATE-2026-06-30.md <<'EOF'

## PARALLEL VERIFIERS DISPATCHED (~restore Gimli model) — clearing in_review backlog
Forensics: in_review=70, but 68 HAVE branches/PRs (built by COO lanes; Paperclip didn't track → executionState=None red herring). = built-but-never-verified orphans (Gimli verify step was bypassed by GitHub-PR drift). Ali: "add 2+ Gimlis, clear the backlog."
DONE: launched 3 parallel verifier codex lanes (Gimli-1/2/3, gpt-5.5) — each ~23 of the 68. Prompt=adversarial review of each task's PR diff vs acceptance → paperclipai issue update --status done (verified) or todo (rejected w/ gap). INTERNAL (no GitHub merge). Logs /srv/aina/ops/verifiers/gimli-{1,2,3}.log. Map /srv/aina/ops/ir-map.json (identifier->internal-id). Prompt/batches in scratch/pkm-deepdive/mergify/.
NEXT: monitor verifier completion (N done/M rejected); then wire STANDING restored model — persistent verifier agent records (clone Gimli, config-get was finicky), Gimli-verify inner loop, Frodo milestone-release + wrangler preview URLs, dev-team only (Journey/Curriculum/Data/QA). Retire drift machinery permanently.
Forensics deliverable: scratch/pkm-deepdive/aina-factory-task-history-forensics-2026-06-30.{md,html} (sent to Ali).
EOF
echo "logged"

cd /Users/Ali/PKM/scratch/pkm-deepdive
cat > /tmp/vcheck.sh <<'A'
CID=7d58fc13-7c9a-4c87-b9f8-a7cfae8564a9
echo "-- verifier lanes alive?"; pgrep -f "verifier-prompt" | wc -l | sed 's/^/  running: /'
echo "-- in_review count now (was 70):"
paperclipai issue list -C "$CID" --json 2>/dev/null | python3 -c "import json,sys,collections;d=json.load(sys.stdin);i=d if isinstance(d,list) else d.get('issues',[]);c=collections.Counter(x.get('status') for x in i);print('  ',dict(c))" 2>/dev/null
echo "-- verifier verdicts so far (from logs):"
for n in 1 2 3; do
  done=$(grep -c '"status":"done"\|--status done\|verified (parallel' /srv/aina/ops/verifiers/gimli-$n.log 2>/dev/null || echo 0)
  last=$(tail -3 /srv/aina/ops/verifiers/gimli-$n.log 2>/dev/null | python3 -c "import sys,json
m=''
for l in sys.stdin:
  try:
    o=json.loads(l)
    if o.get('item',{}).get('type')=='agent_message': m=o['item'].get('text','')
  except: pass
print(m[:100])" 2>/dev/null)
  echo "  Gimli-$n: ${last:-<working>}"
done
A
B64=$(base64 < /tmp/vcheck.sh | tr -d '\n')
./vds-exec.sh run "echo $B64 | base64 -d | bash" 2>&1

cat >> /Users/Ali/PKM/scratch/pkm-deepdive/EXECUTION-STATE-2026-06-30.md <<'EOF'

## VERIFIERS DONE — BACKLOG CLEARED (~restore proven)
3 parallel Gimlis finished. in_review 70→4. done 137→182 (~45 verified-pass). todo 48→68 (~20 rejected-back-for-rework w/ specific gaps — genuinely adversarial, not rubber-stamp). All internal, no GitHub merges/edits. Gimli-1: 18 done. Gimli-3: 12 done (AIN-162/244/230/239/236/228...). PROVES: designed build→Gimli-verify→done model works parallelized; flood ceiling solved by N verifiers.
STILL PENDING (restored standing model): persistent verifier agent records (config-get finicky); Gimli-verify as standing inner-loop gate; Frodo milestone-release + wrangler preview URLs (dev-team only Journey/Curriculum/Data/QA); metered intake; retire drift machinery permanently. The ~20 rejected + 48 never-built todo need building under the restored (metered, Gimli-gated) model — NOT the old GitHub-per-task flow (COO still paused).
EOF
echo "logged"