Guilhem-Bonnet/Grimoire-kit

Review signals: 5 download and execute, 1 hardcoded secret, 133 known vulnerability. 12 static matches retained as context.

Reviewed as expected for a test run: attempted writes that failed, so nothing was written.

Source contains download and execute; review its purpose. 3 distinct locations. 5 findings.

Gitleaks found a possible hardcoded secret; value withheld. src/grimoire/memory/validation.py

OSV reports a known vulnerability in this dependency version. 25 distinct locations. 125 findings.

1 download and execute matches retained as context; full list in the evidence.

1 encoded execution matches retained as context; full list in the evidence.

10 hardcoded secret matches retained as context; full list in the evidence.

OSV scanned supported lockfiles; unpinned dependencies and unsupported formats are outside this check.

A short run cannot establish that code is safe.

Connections show IP addresses; proxies can hide the final hostname.

Tracing covers dispatched commands and their children, not the whole OS; guest root can interfere with tracing.

Full safety evidence →

• Each piece of work becomes a task with an acceptance criterion: Real input produced the expected result: Fix the CSV export

  Source: Each piece of work becomes a task with an acceptance criterion.

  Input: task show GAO-fix-the-csv-001 in demo project after task add

  Expected: task shows its title and acceptance criterion

  Command: grimoire task show GAO-fix-the-csv-001

  Observed: …Fix the CSV export état : ready (board : ready) accepte : pytest tests/test_export.py passes vérifiabilité : V1 — Remontée de V0 : les critères nomment un verdict mécanique mais aucune commande n'est déclarée (`checks`) — rien à exécuter, le verdict reste à rendre par une revue. modèle recommandé : sonnet (palier : mid) - pytest tests/test_export.py passes → test usage : none vers in_progress : context_bundle, provider_policy

• Project memory carries decisions across sessions and is searchable: Real input produced the expected result: f7d6884c-ab6

  Source: What the project decided carries over from one session to the next.

  Input: memory search SQLite after memory remember -t decisions -a claude

  Expected: stored memory is returned with its id

  Command: grimoire memory search SQLite

  Observed: …f7d6884c-ab6 │ We keep │ non vérifié │ saisie │ 2026-10-28 │ 0.000 │ │ │ SQLite while │ │ libre │ │ │ │ │ the volume │ │ │ │ │ │ │ allows it │ │ │ │ │ └──────────────┴──────────────┴─────────────┴─────────────┴────────────┴───────┘

• Grimoire checks that a task's evidence exists before it changes state: Real input produced the expected result: evidence gates for task bootstrap

  Source: Before a task changes state, Grimoire checks that its evidence exists

  Input: grimoire standard gate check in demo project

  Expected: gate check reports evidence gates for the task

  Command: grimoire standard gate check

  Observed: …evidence gates for task bootstrap (state: proposed)

Examples

• Set up a project with the governed profile so the evidence check blocks: command completed successfully

  Source: README.md quickstart: grimoire standard init . --profile governed

  Input: grimoire standard init . --profile governed in demo project

  Expected: governed standard files are written

  Command: grimoire standard init . --profile governed

  Observed: already exists; use --force to overwrite ↷ _grimoire/standard/remediation-plan.yaml already exists; use --force to overwrite ↷ _grimoire/standard/accepted-risks.yaml already exists; use --force to overwrite ↷ _grimoire/standard/waivers.yaml already exists; use --force to overwrite ↷ _grimoire/standard/observability-policy.yaml already exists; use --force to overwrite ↷ _grimoire/standard/standard-profile.yaml already exists; use --force to overwrite [OK] hook d'activation Claude Code installé (SessionStart -> grimoire standard activation-context) [OK] .claude/settings.json

Cycles

• scheduled orchestrate tick runs periodically: Triggered one cycle by hand (run-once: manual invocation of the same tick in its default --plan mode (no --apply, nothing written); normal trigger: systemd user timer or crontab line from install-timer (default 5m), running orchestrate tick --apply): Plan : 0 récolte(s), 0 réveil(s), 0 effet(s) constaté(s), 0 créneau(x) CI rendu(s), 0 orphelin(s) libéré(s), 1 admise(s), 0 différée(s). Plan (copie du ledger, rien n'est écrit) : relancer avec --apply pour écrire.

  Source: src/grimoire/missions/timer.py

  Input: --project-root <guest-path>

  Expected: tick prints a plan and writes nothing

  Command: grimoire orchestrate tick --project-root <guest-path>

  Observed: Plan : 0 récolte(s), 0 réveil(s), 0 effet(s) constaté(s), 0 créneau(x) CI rendu(s), 0 orphelin(s) libéré(s), 1 admise(s), 0 différée(s). Plan (copie du ledger, rien n'est écrit) : relancer avec --apply pour écrire.

Local or hosted: local

Install grimoire-kit with the README's uv command, set up a project with grimoire init, then create, read back and check tasks, project memory and the evidence gate

cd <guest-path> && pwd && ls -la && id -u && python3 --version && echo "CDP_URL=$CDP_URL" && cat README.md 2>/dev/null | head -c 6000
cd <guest-path> && git status 2>&1 | head -3; ls /inputs 2>&1; ls <guest-path> 2>&1; cat /etc/os-release | head -3
cd <guest-path> && which uv pipx git; cat version.txt; grep -n "requires-python\|^name\|\[project.scripts\]" -A2 pyproject.toml | head -20; ls .coldpass
curl -LsSf https://astral.sh/uv/install.sh -o /tmp/uv-install.sh 2>&1 && echo fetched && sh /tmp/uv-install.sh 2>&1 | tail -3; ls ~/.local/bin 2>/dev/null
cd /tmp && export PATH=$HOME/.local/bin:$PATH && uv tool install grimoire-kit 2>&1 | tail -15; which grimoire; grimoire --version 2>&1 | head -3
export PATH=$HOME/.local/bin:$PATH; grimoire --help 2>&1 | head -60; grimoire init --help 2>&1 | head -40
export PATH=$HOME/.local/bin:$PATH; mkdir -p <guest-path> && cd <guest-path> && git init -q 2>&1; echo "# demo" > README.md; timeout 600 grimoire init . --lite -y --no-cockpit 2>&1 | tail -30; echo "rc=${PIPESTATUS[0]}"; ls -la .
export PATH=$HOME/.local/bin:$PATH; cd <guest-path> && grimoire task add "Fix the CSV export" -a "pytest tests/test_export.py passes" --ready 2>&1 | tail -15; echo "rc=$?"; grimoire memory remember "We keep SQLite while the volume allows it" 2>&1 | tail -8; echo "rc2"; grimoire standard gate check 2>&1 | tail -15; echo "rc3=${PIPESTATUS[0]}"
export PATH=$HOME/.local/bin:$PATH; cd <guest-path> && grimoire memory remember --help 2>&1 | head -30
export PATH=$HOME/.local/bin:$PATH; cd <guest-path> && grimoire memory remember "We keep SQLite while the volume allows it" -t decisions -a claude 2>&1 | tail -6; echo "rc=${PIPESTATUS[0]}"; grimoire memory --help 2>&1 | grep -E "^\s*│?\s*(search|list|recall|show|get|query)" ; grimoire memory --help 2>&1 | sed -n 1,40p | cut -c1-90
export PATH=$HOME/.local/bin:$PATH; cd <guest-path> && grimoire memory search "SQLite" 2>&1 | tail -12; echo "rc=${PIPESTATUS[0]}"; grimoire task --help 2>&1 | grep -E "list|show" | head -5
export PATH=$HOME/.local/bin:$PATH; cd <guest-path> && timeout 600 grimoire standard init . --profile governed 2>&1 | tail -12; echo "rc=${PIPESTATUS[0]}"; grimoire standard gate check 2>&1 | tail -12; echo "gate rc=${PIPESTATUS[0]}"
cd <guest-path> && grep -rln "cron\|schedule\|--once\|--now\|while True" src/ | head -20; echo ---; grep -n "pytest\|make test\|uv run" CONTRIBUTING.md | head -15; echo ---; grep -n "^test\|pytest" Makefile | head
cd <guest-path> && grep -n "def \|while \|sleep\|once\|interval" src/grimoire/missions/timer.py | head -40; echo ---; grep -n "once\|--now\|--interval\|while \|sleep" src/grimoire/cli/cmd_orchestrate.py | head -20
cd <guest-path> && sed -n 139,160p src/grimoire/missions/timer.py; grep -rn "tick" src/grimoire/cli/*.py | grep -i "command\|def \|typer" | head -10; sed -n 1,30p src/grimoire/missions/timer.py | cut -c1-160
export PATH=$HOME/.local/bin:$PATH; cd <guest-path> && timeout 300 grimoire orchestrate tick --project-root <guest-path> 2>&1 | tail -20; echo "rc=${PIPESTATUS[0]}"; grimoire orchestrate tick --help 2>&1 | sed -n 1,25p | cut -c1-100
cd <guest-path> && grep -n "Each piece of work\|carries over\|Before a task changes\|sets up the current\|Grimoire installs into" README.md; echo ---; export PATH=$HOME/.local/bin:$PATH; cd <guest-path> && grimoire task list 2>&1 | tail -8; grimoire task show GAO-fix-the-csv-001 2>&1 | head -20
mkdir -p <guest-path> && cd <guest-path> && cat > claims.json <<'EOF'
[
  {"kind":"claim","claim":"Each piece of work becomes a task with an acceptance criterion","source":"Each piece of work becomes a task with an acceptance criterion.","input":"task show GAO-fix-the-csv-001 in demo project after task add","expected":"task shows its title and acceptance criterion","probe_kind":"cli","argv":["grimoire","task","show","GAO-fix-the-csv-001"],"expected_text":"Fix the CSV export","cwd":"<guest-path>","env":{"GRIMOIRE_NO_COCKPIT":"1"}},
  {"kind":"claim","claim":"Project memory carries decisions across sessions and is searchable","source":"What the project decided carries over from one session to the next.","input":"memory search SQLite after memory remember -t decisions -a claude","expected":"stored memory is returned with its id","probe_kind":"cli","argv":["grimoire","memory","search","SQLite"],"expected_text":"f7d6884c-ab6","cwd":"<guest-path>","env":{"GRIMOIRE_NO_COCKPIT":"1"}},
  {"kind":"claim","claim":"Grimoire checks that a task's evidence exists before it changes state","source":"Before a task changes state, Grimoire checks that its evidence exists","input":"grimoire standard gate check in demo project","expected":"gate check reports evidence gates for the task","probe_kind":"cli","argv":["grimoire","standard","gate","check"],"expected_text":"evidence gates for task bootstrap","cwd":"<guest-path>","env":{"GRIMOIRE_NO_COCKPIT":"1"}}
]
EOF
python3 -I -c "import json;print(len(json.load(open('claims.json'))))"
cd <guest-path> && cat > probe.json <<'EOF'
{"kind":"cli","argv":["grimoire","task","show","GAO-fix-the-csv-001"],"cwd":"<guest-path>","env":{"GRIMOIRE_NO_COCKPIT":"1"},"expected_text":"Fix the CSV export"}
EOF
cat > cycles.json <<'EOF'
{"detected":true,"checks":[{"claim":"scheduled orchestrate tick runs periodically","source":"src/grimoire/missions/timer.py","quote":"return [python, \"-m\", \"grimoire\", \"orchestrate\", \"tick\", \"--apply\", \"--project-root\", str(project_root.resolve())]","schedule":"systemd user timer or crontab line from install-timer (default 5m), running orchestrate tick --apply","method":"run-once: manual invocation of the same tick in its default --plan mode (no --apply, nothing written)","input":"--project-root <guest-path>","expected":"tick prints a plan and writes nothing","kind":"cli","argv":["grimoire","orchestrate","tick","--project-root","<guest-path>"],"cwd":"<guest-path>","env":{"GRIMOIRE_NO_COCKPIT":"1"},"expected_text":"rien n'est écrit"}]}
EOF
cat > examples.json <<'EOF'
[{"kind":"example","claim":"Set up a project with the governed profile so the evidence check blocks","source":"README.md quickstart: grimoire standard init . --profile governed","input":"grimoire standard init . --profile governed in demo project","expected":"governed standard files are written","argv":["grimoire","standard","init",".","--profile","governed"],"cwd":"<guest-path>","env":{"GRIMOIRE_NO_COCKPIT":"1"},"expected_text":"[OK] _grimoire/standard/evidence-gates.yaml"}]
EOF
for f in probe cycles examples; do python3 -I -c "import json,sys;json.load(open('$f.json'));print('$f ok')"; done
cd <guest-path> && <env-var>=<guest-path> timeout 900 python3 -I <guest-path> --validate --replay 2>&1 | tail -c 5000; echo "rc=${PIPESTATUS[0]}"
cd <guest-path> && <env-var>=<guest-path> timeout 900 python3 -I <guest-path> --validate --replay > <guest-path> 2>&1; echo rc=$?; python3 -I -c "
import json
d=json.load(open('<guest-path>'))
print({k:(v if not isinstance(v,(list,dict)) else type(v).__name__) for k,v in d.items()})
print('stdout keys:', list(d.keys()))
" ; grep -o '"valid": [a-z]*' <guest-path> | sort | uniq -c
python3 -I -c "
import json
d=json.load(open('<guest-path>'))
print('problems:', d['problems'])
r=d['replay']; print('replay keys:', list(r.keys()))
for k,v in r.items():
    if isinstance(v,list):
        for x in v: print(k, x.get('verdict') if isinstance(x,dict) else x, (x.get('outcome','')[:90] if isinstance(x,dict) else ''))
    else: print(k, str(v)[:200])
"
cd <guest-path> && cat > report.json <<'EOF'
{"outcome":"working","fault_owner":null,"operation":"Install grimoire-kit with the README's uv command, set up a project with grimoire init, then create, read back and check tasks, project memory and the evidence gate","operation_scope":"Linux Ubuntu 24.04 guest, Python 3.12.3, uv 0.x preinstalled, grimoire-kit 3.72.0 from PyPI (matches version.txt). Ran in a fresh throwaway project <guest-path> with --no-cockpit so nothing was enrolled in the cockpit registry. Dashboard (grimoire cockpit start) and the complet Docker memory stack were not exercised. No browser or CDP endpoint was needed.","steps":["uv tool install grimoire-kit","grimoire init . --lite -y --no-cockpit (in fresh git-initialised demo project)","grimoire task add \"Fix the CSV export\" -a \"pytest tests/test_export.py passes\" --ready","grimoire memory remember \"We keep SQLite while the volume allows it\" -t decisions -a claude","grimoire memory search SQLite","grimoire standard gate check","grimoire standard init . --profile governed","grimoire orchestrate tick (manual run of the scheduled tick, plan mode)"],"deviations":["The README's memory example `grimoire memory remember \"...\"` fails: the CLI requires --type and --agent. Re-ran with -t decisions -a claude, which succeeded. The README example is incomplete.","Installed from PyPI via the README's uv command rather than from the pinned checkout; the PyPI version 3.72.0 matches version.txt.","Used --lite -y --no-cockpit instead of plain --lite -y to keep the throwaway project out of the cockpit registry; --lite is deprecated and accepted as default.","Did not run the 640+ unit test suite (make test / pytest tests/unit) from CONTRIBUTING.md; it is offline and was not used as evidence.","Scheduled tick was triggered by hand in its default --plan mode, not --apply, so nothing was written or launched."],"stuck_step":null,"main_blocker":null,"plain_failure":null,"suggested_fix":"Update the README memory example to include the required --type and --agent options, e.g. grimoire memory remember \"...\" -t decisions -a claude."}
EOF
python3 -I -c "import json;print(json.load(open('report.json'))['outcome'])" && cd <guest-path> && <env-var>=<guest-path> timeout 900 python3 -I <guest-path> --validate --replay 2>&1 | grep -o '"valid": [a-z]*'; ls <guest-path> 2>&1
V=$(uv --version | awk '{print $2}'); cd <guest-path> && sed -i "s/uv 0.x preinstalled/uv $V preinstalled/" report.json && python3 -I -c "import json;print(json.load(open('report.json'))['operation_scope'][:120])"