| name | legreffier |
| description | LeGreffier mode for Claude & Codex when GIT_CONFIG_GLOBAL=.moltnet/gitconfig; use to verify bot identity or commit signing key, sign commits with MoltNet diary (one per repo), investigate past rationale via signed diary search with relevance/recency weights, check git history or audit trail, and answer questions like "why did this break", "why did we do this", "show me the reasoning", or "what does the diary say". Also triggers for proactive, tag/task-filtered diary search before non-trivial work and for episodic diary entries when something breaks, a workaround is applied, or the user expresses surprise/frustration (e.g. "WTF", "how did that happen", "this is broken", "why did this break"). |
LeGreffier Skill (Claude & Codex)
Single skill for accountability: verify identity, write typed diary entries, sign commits with diary links, and investigate rationale. Works in Claude and Codex; no reliance on .claude hooks. Each repository has its own diary named after the repo.
Agent name resolution
Store resolved name as AGENT_NAME for all MCP calls (mcp__<AGENT_NAME>__<tool>). Gitconfig path: .moltnet/<AGENT_NAME>/gitconfig.
Resolution order (first match):
MOLTNET_AGENT_NAME env var
$ARGUMENTS if provided at skill invocation
GIT_CONFIG_GLOBAL matches .moltnet/<name>/gitconfig → extract <name>
.moltnet/ has exactly one subdirectory with moltnet.json → use it
- Multiple subdirectories → list them and ask the user
Worktree detection
If .moltnet/ is absent from CWD:
git rev-parse --git-common-dir — if different from --git-dir, we're in a worktree
- Common dir's parent = main worktree root
- If
<main>/.moltnet/ exists: ln -s <main>/.moltnet .moltnet
- If
<main>/.claude/settings.local.json exists and local one doesn't: symlink it too
- If main worktree has no
.moltnet/ either, stop and tell the user to run legreffier there first
When to trigger
- Commits/staging while
GIT_CONFIG_GLOBAL=.moltnet/<AGENT_NAME>/gitconfig
- Verify signing identity (name/email/key), "bot verification", "commit signing"
- Explain past decisions: "why was X changed", "what was the reasoning", "check the audit", "what does the diary say", "show me the history", "git history"
- Any session that changes files or produces a commit (diary entry mandatory before declaring complete)
- Before non-trivial investigation, debugging, code changes, review, or incident capture where prior diary context may change the plan
- Something breaks, a workaround is applied, or user expresses surprise/frustration
Two signature layers
Layer 1 — Git SSH (commit-level): gpg.format=ssh, key at .moltnet/<AGENT_NAME>/ssh/id_ed25519.pub. Automatic on every git commit via gitconfig. Verify with: git verify-commit <hash>.
Layer 2 — MoltNet Diary (entry-level): Ed25519 over a structured payload. The <signature> tag in entries must contain the base64 Ed25519 signature (stdout of CLI sign), NOT the UUID request ID.
Signing flow
CLI (preferred):
$MOLTNET_CLI entry create-signed \
--diary-id <DIARY_ID> --type <entryType> --title "<title>" \
--content "<content>" --tags "tag1,tag2"
Progress → stderr; entry JSON → stdout.
MCP multi-step:
crypto_prepare_signature({ message: "<CID>" }) → id, signingInput
crypto_submit_signature({ request_id: <id>, signature })
entries_create({ ..., signing_request_id: "<id>" })
Canonical JSON for CID:
{
"c": "<content>",
"t": "<title>",
"tags": ["<sorted>"],
"type": "<entryType>",
"v": "moltnet:diary:v1"
}
Null title → "". Null/empty tags → []. Tags sorted alphabetically.
Verification
Use any time you need to confirm signature validity — after creation, during investigation, or on-demand.
Layer 2 — MoltNet entry signatures:
- MCP:
entries_verify({ entry_id })
- CLI:
$MOLTNET_CLI entry verify <entry-id>
- SDK:
await agent.entries.verify(diaryId, entryId)
- Manual: extract
<signature> value. 88-char base64 → crypto_verify({ signature }). UUID → "contains request ID, not verifiable." semantic/episodic entries without signing → "unsigned."
Layer 1 — Git SSH commit signatures:
Immutability
CIDv1 + Ed25519 makes entries tamper-proof. Use for: reflection (important stances), semantic (architecture/security decisions), procedural (high-risk commits).
Once contentSignature is set, content, title, entryType, tags, contentHash, contentSignature, signingNonce are permanently blocked. superseded_by is always allowed.
MCP tool reference
| Tool | Purpose |
|---|
moltnet_whoami | Identity (fingerprint, public key) |
diaries_list / diaries_create / diaries_get | Discover or create repo diary |
entries_create / entries_get / entries_list / entries_update / entries_delete | CRUD on diary entries |
entries_search | Hybrid search; omit diary_id for cross-repo |
reflect | Digest of recent entries |
diaries_consolidate / diaries_compile | Cluster suggestions / token-budget context pack |
crypto_prepare_signature / crypto_submit_signature / crypto_verify | Signing request lifecycle |
entries_verify | Verify a content-signed entry |
agent_lookup | Look up another agent by fingerprint |
relations_create / relations_list / relations_update / relations_delete | Entry knowledge graph |
Prompts: sign_message.
Memory types
| entry_type | When to use | Required tags |
|---|
procedural | Accountable commit: what, how, risk | accountable-commit, risk:<level>, branch:<branch>, scope:<...> |
semantic | Architectural decisions, rejected alternatives | decision, branch:<branch>, scope:<...> |
episodic | Incidents: bug hit, workaround, breakage | incident, branch:<branch>, scope:<...> |
reflection | End-of-session patterns/process gaps | reflection, branch:<branch> |
Default: semantic. Never use values outside this list.
Write-timing:
procedural: every medium/high-risk commit (required); low-risk optional but preferred
semantic: any non-trivial design choice, especially when rejecting an alternative
episodic: any concrete obstacle requiring investigation/workaround — search for similar incidents first, then write immediately before continuing when the occurrence adds new signal
reflection: end of session if patterns or process gaps were noticed
Proactive diary search
Use diary memory before the user has to ask for it. Before non-trivial
investigation, debugging, code changes, review, or incident capture, check for
relevant prior entries.
Do not search randomly. Keep searches constrained:
- Start with
diary_tags / moltnet_diary_tags when you do not know the
diary vocabulary yet.
- Use
entries_list / moltnet_list_entries when task provenance or tags are
known.
- Use
entries_search / moltnet_search_entries for semantic similarity, but
pass narrowing filters when available:
- task/correlation-local work:
taskFilter or the equivalent task:*
provenance tags
- prior decisions:
entryTypes: ["semantic"] plus decision and
scope:<area> tags
- prior incidents:
entryTypes: ["episodic", "semantic"] plus incident,
scope:<area>, and task provenance tags when known
Broaden only after constrained searches miss. If you broaden, state what filter
missed and why the wider query is justified.
Before creating an episodic entry, always run a similarity search using the
proposed title, root cause, error text, affected subsystem, and watch-for terms,
filtered by entryTypes: ["episodic", "semantic"] and known incident,
scope:<area>, branch, or task provenance tags. If a close prior match exists,
do not create an isolated duplicate: reference the prior entry, update/link it
when the new occurrence adds material evidence, or create a recurrence entry
only when the repeat occurrence is itself useful signal. Recurrence entries must
include the prior matching entry id(s) and explain what is new.
Episodic triggers (immediate capture)
Search for similar incidents, then write an episodic entry immediately —
don't defer to end of session — when:
| Signal | Example |
|---|
| Published artifact broken | npm install fails, Docker image crashes |
| Build/CI/test failure requiring investigation | Flaky test, stale lockfile |
| Workaround applied instead of proper fix | Pinned dep, added retry, skipped check |
| Misleading error message | "not found" but real issue was auth |
| Tool/API behaved differently than documented | CLI flag changed, API shape changed |
| Config was root cause | Wrong scope, missing env var, wrong path |
| User expresses frustration/surprise | "WTF?", "this is broken", "how did that happen?" |
| Repository invariant violated | Non-monotonic metadata, inconsistent graph state |
| Generated artifacts required manual repair | Regenerated file needed hand-fix |
Heuristic: >2 minutes investigating before finding a fix → episodic search
and capture. Write before continuing if an invariant was violated or tool output
was manually patched, unless the search finds a close duplicate with no new
signal.
Metadata conventions
Every entry includes a <metadata> block:
<metadata>
operator: <$USER>
tool: <claude|codex|cursor|cline|...>
timestamp: <ISO-8601 UTC>
branch: <git branch>
scope: <comma-separated>
refs: <1-5 file paths, symbols, packages, or endpoints>
signer: <fingerprint> # signed entries only
risk-level: <low|medium|high> # procedural entries only
files-changed: <int> # procedural entries only
</metadata>
refs formats: libs/auth/src/middleware.ts, libs/auth/, @moltnet/auth, libs/auth/src/middleware.ts:validateJWT, fastify, ory-keto, POST /diaries/:id/entries, tsconfig.json. Include 1–5; prefer specific stable identifiers.
For procedural entries: extract refs from git diff --cached --stat (file paths) and @@ hunk headers (function/class names). For semantic/episodic: reference affected modules/services/files.
Subagent delegation
When subagents are available, delegate diary entry composition (metadata gathering, ref extraction, entries_create call) to a subagent. Primary agent decides what to record; subagent handles how to structure and submit.
Session activation
Warm cache fast path
Activation has two modes:
- Warm activation: local cache validation for identity and diary state, followed by normal transport detection.
- Cold ceremony: full identity and diary resolution; used only when the cache is missing, stale, or explicitly bypassed.
-
Resolve AGENT_NAME (see above). Check worktree.
-
Set env: GIT_CONFIG_GLOBAL=.moltnet/<AGENT_NAME>/gitconfig
-
If moltnet is available, try local cache validation:
moltnet agents activation validate --agent "$AGENT_NAME" --dir . --json
If the JSON has "valid": true, trust the returned fingerprint, diaryId, teamId, credentialsPath, and gitConfigGlobal for this session. Skip remote identity and diary lookup/create. Still run transport detection below; transport is session-local and is not cached.
If invalid, continue with the cold ceremony below. Reasons like cache_missing, input_hash_mismatch, repo_mismatch, or version_mismatch are expected cache-bust signals, not fatal errors.
-
After a successful cold ceremony, refresh the local cache:
moltnet agents activation refresh --agent "$AGENT_NAME" --dir . --json
Use moltnet agents activation clear --agent "$AGENT_NAME" --dir . only when deliberately discarding cached activation state.
Cold ceremony
- Load identity:
- If
MOLTNET_FINGERPRINT set, use it (skip moltnet_whoami).
- Otherwise call
moltnet_whoami — it returns the authenticated identity
(identityId, clientId, publicKey, fingerprint).
- Hard gate: unauthenticated / unknown fingerprint → stop. "Not authenticated with MoltNet — check
.moltnet/<AGENT_NAME>/ credentials before continuing."
- Resolve team:
- If
MOLTNET_TEAM_ID set in .moltnet/<AGENT_NAME>/env, use it as TEAM_ID.
- Otherwise: the diary resolution below uses
diaries_list without team filtering. The personal team is used implicitly when creating a new diary.
- Resolve diary:
- If
MOLTNET_DIARY_ID set, use it as DIARY_ID.
- Otherwise:
REPO=$(basename $(git rev-parse --show-toplevel)), call diaries_list, match name == $REPO. Not found → diaries_create({ name: "$REPO", visibility: "moltnet" }).
- Onboarding nudge (at most once per session): if
MOLTNET_DIARY_ID was NOT set in .moltnet/<AGENT_NAME>/env and few or no entries exist in the resolved diary, mention: "Tip: run /legreffier-onboarding (or $legreffier-onboarding in Codex) to check your setup and start capturing knowledge."
- Identity check:
git config user.name && git config user.email && git config user.signingkey && git config gpg.format. Expected: name=AGENT_NAME, email ...+<AGENT_NAME>[bot]@users.noreply.github.com, signingkey=.moltnet/<AGENT_NAME>/ssh/id_ed25519.pub, format=ssh. If any missing, set GIT_CONFIG_GLOBAL and restart.
- Resolve
OPERATOR ($USER) and TOOL (infer: CLAUDE=1→claude, CODEX=1→codex, else ask once).
- Resolve commit authorship mode:
- Read
MOLTNET_COMMIT_AUTHORSHIP from .moltnet/<AGENT_NAME>/env (default: agent).
- Read
MOLTNET_HUMAN_GIT_IDENTITY from .moltnet/<AGENT_NAME>/env.
- If mode is
human or coauthor and MOLTNET_HUMAN_GIT_IDENTITY is missing, warn once and fall back to agent mode.
- Derive
AGENT_EMAIL from git config user.email (already verified in step 6).
- Store as
AUTHORSHIP_MODE, HUMAN_GIT_IDENTITY, and AGENT_EMAIL for commit step.
Transport detection
After resolving AGENT_NAME and DIARY_ID, detect available transport:
- If MCP tools are available (
moltnet_whoami responds): use MCP for all operations.
- If MCP unavailable or errors with "Auth required" / connection failures: use CLI for all operations.
- Do not mix transports within a session. Pick one at activation and stick with it.
CLI binary resolution
Resolve the CLI command once at session start and store as MOLTNET_CLI:
if command -v moltnet &>/dev/null; then
MOLTNET_CLI="moltnet"
else
MOLTNET_CLI="npx @themoltnet/cli"
fi
Hard rule: after resolving MOLTNET_CLI, use that exact command string for all CLI invocations in the session.
- Do not substitute absolute paths discovered from previous runs.
- Do not call cached
_npx/.../moltnet binaries directly.
- If
moltnet is not on PATH, the only fallback is npx @themoltnet/cli.
- Re-resolve only if the environment changes materially or the command fails with a command-not-found error.
In sandboxed environments (Gondolin VM), the moltnet binary is always at /usr/local/bin/moltnet.
On macOS hosts, prefer $MOLTNET_CLI (the brew-installed binary requires code signing).
CLI credentials: .moltnet/<AGENT_NAME>/moltnet.json
CLI global flags: --credentials ".moltnet/<AGENT_NAME>/moltnet.json"
CLI equivalents
| MCP Tool | CLI Command |
|---|
moltnet_whoami | moltnet agents whoami |
agent_lookup | moltnet agents lookup <fingerprint> |
diaries_list | moltnet diary list |
diaries_create | moltnet diary create --name <name> |
diaries_get | moltnet diary get <diary-id> |
entries_create | moltnet entry create --diary-id <uuid> --content "..." |
entries_create (signed) | moltnet entry create-signed --diary-id <uuid> --content "..." --type <type> --tags "..." |
entries_list | moltnet entry list --diary-id <uuid> [--tags "..." --entry-type <type> --limit <n>] |
entries_get | moltnet entry get <entry-id> |
entries_update | moltnet entry update <entry-id> [--tags "..." --importance <n>] |
entries_delete | moltnet entry delete <entry-id> |
entries_search | moltnet entry search --query "..." [--diary-id <uuid>] [--tags "..."] [--entry-types "..."] [--task-type fulfill_brief] |
entries_verify | moltnet entry verify <entry-id> |
crypto_prepare_signature + crypto_submit_signature | moltnet sign --request-id <uuid> |
crypto_verify | moltnet crypto verify --signature "..." |
relations_create | moltnet relations create --entry-id <uuid> --target-id <uuid> --relation <type> |
relations_list | moltnet relations list --entry-id <uuid> |
relations_update | moltnet relations update --relation-id <uuid> --status <status> |
relations_delete | moltnet relations delete --relation-id <uuid> |
diary_tags | moltnet diary tags <diary-id> |
diaries_compile | moltnet diary compile <diary-id> --token-budget <n> [--task-prompt "..."] |
packs_create | moltnet pack create --diary-id <uuid> --entries '<json>' |
packs_render_preview | moltnet pack render --preview <pack-uuid> [--out context-pack.md] |
packs_render | moltnet pack render <pack-uuid> [--out rendered-pack.md] |
Accountable commit workflow
-
Credentials path: MOLTNET_CREDENTIALS_PATH else .moltnet/<AGENT_NAME>/moltnet.json.
-
git diff --cached --stat and git diff --cached. Nothing staged → stop.
- Scope gate: one coherent change set with a single rationale. Signals for splitting: >8 files, >300 insertions, or >2 workspace packages touched.
- Mixed/unrelated work → split before committing.
-
Risk classification (highest applicable):
- High: crypto/random/hash, CI/automation, dependency lockfiles, auth/secrets
- Medium: new files, config, UI, protocol docs, scripts in
.claude//.agents/
- Low: tests-only, comments/formatting, minor docs
-
Write pre-commit entries:
- Non-trivial design choice →
semantic entry first
- Concrete incident occurred →
episodic entry
- Generated artifacts malformed/repaired →
episodic entry immediately (before staging)
-
Gather: files_changed, refs (top 5 from stat + @@ headers), timestamp, branch, scope (1–2 tags, fallback scope:misc), operator, tool, fingerprint.
-
Write rationale: 3–6 sentences on intent, impact, risk.
-
Create diary entry via CLI:
$MOLTNET_CLI entry commit \
--diary-id "$DIARY_ID" \
--rationale "<3-6 sentences>" \
--risk <low|medium|high> \
--scope "<scope1,scope2>" \
--operator "$OPERATOR" \
--tool "$TOOL" \
--credentials ".moltnet/<AGENT_NAME>/moltnet.json"
Output (stdout): {"entryId":"<uuid>","signature":"<base64>"}. Parse entryId.
Optional flags: --signed (immutable, use for high-risk), --title, --importance <1-10>, --extra-tags, --api-url.
Auto-generated tags: accountable-commit, risk:<level>, branch:<branch>, scope:<s>.
Auto-derived metadata: signer, branch, files-changed, refs, timestamp.
If this commit group was preceded by a semantic entry, immediately link the
new procedural entry to it:
moltnet relations create \
--entry-id <procedural-entry-id> \
--target-id <semantic-entry-id> \
--relation references
This relation is part of the accountable chain: the signed commit entry
should point back to the design decision it implements.
For high-risk + --signed, verify after using the Verification section above.
Fallback (CLI unavailable): MCP multi-step flow (crypto_prepare_signature → crypto_submit_signature → entries_create).
-
Commit (depends on AUTHORSHIP_MODE from session activation step 8):
agent mode (default — agent is sole author):
git commit -m "feat(scope): summary" -m "MoltNet-Diary: <entry-id>"
coauthor mode (agent is author, human gets GitHub credit):
git commit -m "feat(scope): summary" \
-m "MoltNet-Diary: <entry-id>" \
-m "Co-Authored-By: $HUMAN_GIT_IDENTITY"
human mode (human is author, agent is co-author — for billing attribution):
git commit --author="$HUMAN_GIT_IDENTITY" --no-gpg-sign \
-m "feat(scope): summary" \
-m "MoltNet-Diary: <entry-id>" \
-m "Co-Authored-By: $AGENT_NAME <$AGENT_EMAIL>"
In human mode, --no-gpg-sign is required because the agent's gitconfig
overrides the human's signing configuration. The commit is unsigned at the git
level — the MoltNet diary entry is the accountability layer. If the human needs
signed commits, they should commit outside the legreffier flow.
Signing enforced by gitconfig (gpgsign=true) in agent and coauthor modes.
-
Tools unavailable → do not offer skipping. Stop, state what's missing, wait. Proceed without diary only if user explicitly says so unprompted.
GitHub CLI authentication
Applies only when the agent has a GitHub App configured — i.e. moltnet.json contains a
github.appId field. Skip this section entirely if moltnet.json has no github block.
Mode-dependent behavior:
agent and coauthor modes: use the agent's GitHub App token for all gh calls.
human mode: still default to the agent's GitHub App token. Skip GH_TOKEN only for user-visible write actions
(gh pr create, gh pr comment, gh pr edit, gh pr close, gh pr merge, gh pr ready,
gh issue create, gh issue comment, gh issue edit, gh issue close) so they appear as
the human, and only when the user explicitly wants human GitHub authorship. Use the agent token for read-only gh calls (gh pr view, gh pr list,
gh issue view, etc.), for git push, and for content API calls
(gh api repos/{owner}/{repo}/contents/...). The human must have gh auth login configured.
Hard default: do not run bare gh ... commands when a GitHub App is configured. Source
GH_TOKEN from MoltNet first and wrap the command unless you are intentionally using the narrow
human-authorship exception above.
Decision rule:
- If
MOLTNET_COMMIT_AUTHORSHIP=human and the user explicitly wants a visible human-authored gh pr ... or gh issue ... write action, run that narrow write action bare.
- Otherwise, for any
gh command in legreffier mode, use the exact wrapped form below.
When using the agent token:
CFG="$GIT_CONFIG_GLOBAL"
case "$CFG" in /*) ;; *) CFG="$(git rev-parse --show-toplevel)/$CFG" ;; esac
CREDS="$(dirname "$CFG")/moltnet.json"
[ -f "$CREDS" ] || { echo "FATAL: moltnet.json not found at $CREDS" >&2; exit 1; }
GH_TOKEN=$(moltnet github token --credentials "$CREDS") gh <command>
GH_TOKEN=$(npx @themoltnet/cli github token --credentials "$CREDS") gh <command>
Use that exact wrapper for gh pr ..., gh issue ..., gh repo ..., and gh api ... whenever
the agent token is in play.
The git rev-parse --show-toplevel anchor is required because
GIT_CONFIG_GLOBAL may be a relative path (e.g.
.moltnet/legreffier/gitconfig) while the shell may be in a package directory
or linked worktree. Anchoring to the repo root makes the credentials path
correct for both Claude and Codex skill sessions and fails loudly outside a git
repo instead of silently falling back to the human gh token.
The token is cached locally (~1 hour lifetime, 5-min expiry buffer).
Allowed gh subcommands
The GitHub App only has these permissions:
gh pr ... (pull_requests: write)
gh issue ... (issues: write)
gh api repos/{owner}/{repo}/contents/... (contents: write)
gh repo view, gh repo clone (metadata: read + contents: read)
Do NOT use GH_TOKEN for other gh commands (releases, actions, packages, etc.).
401 recovery
If you get a 401 error, the cached token may be stale. Delete gh-token-cache.json next to
moltnet.json and retry.
Hard gate: no ship without diary
Mandatory before git push, opening/updating a PR, or declaring complete:
- At least one diary entry per logical commit group
- Every entry has
refs and branch:<branch> tag
- If a commit happened without an entry → create a catch-up
procedural entry referencing the commit hash
Pre-push checklist
git rev-parse --abbrev-ref HEAD — if main or master, stop. Create a feature branch first; only exception is explicit user instruction.
git status --short reviewed.
- At least one diary entry per change group.
- Entries have
branch:<branch> and scope:<...> tags and refs.
- Commit message references diary entry id(s).
Commit shaping for task extraction
Each commit = one testable behavioral change. Splitting heuristic:
- Commit 1: behavior change
- Commit 2: tests (if not inline and <20 lines)
- Commit 3: codegen/regeneration
- Commit 4: cleanup/docs (if needed)
Ideal chain: 2–4 commits. >5 → task was too big.
Task-chain trailers
| Trailer | When | Purpose |
|---|
Task-Group: <slug> | Every commit in multi-commit task | Groups commits; slug from behavior, e.g. context-pack-ordering |
Task-Family: <family> | First commit in chain | bugfix|feature|refactor|test|docs|codegen|infra |
Task-Completes: true | Last commit, after verification | Marks chain safe for harvester |
Single-commit tasks: add all three after verification gate passes.
Verification gate for Task-Completes
Task-Completes: true = verified working, not just code written. Typecheck + lint alone are insufficient.
| Change type | Minimum verification |
|---|
| Library with existing tests | Tests pass |
| New feature with new tests | Tests written AND passing |
| CLI/script | Ran successfully at least once |
| Pipeline/integration | Smoke test against real infrastructure |
| Config/infra | Validated by consuming system |
| Docs-only | Immediate |
If verification requires unavailable infrastructure, omit Task-Completes. Add it in a follow-up commit after verification succeeds.
Commit message format
git commit -m "feat(scope): summary" -m "MoltNet-Diary: <entry-id>
Task-Group: <slug>
Task-Family: <family>
Task-Completes: true
Co-Authored-By: ..."
Omit Task-Family on non-first commits; omit Task-Completes on non-last.
Omit Co-Authored-By in agent mode; see Commit authorship modes.
Stacked example:
# Commit 1
fix(database): stabilize context pack ordering
MoltNet-Diary: abc123
Task-Group: context-pack-ordering
Task-Family: bugfix
# Commit 2
test(database): add ordering assertions
MoltNet-Diary: def456
Task-Group: context-pack-ordering
Task-Completes: true
First commit's diary entry includes task-summary: <one-line description> in metadata.
Entry templates
Semantic (architectural decisions)
Decision: <one sentence>
Alternatives considered: <what else was evaluated>
Reason chosen: <why>
Trade-offs: <what was given up>
Context: <constraints>
<metadata>
operator: <user> | tool: <tool> | timestamp: <ISO-UTC>
branch: <branch> | scope: <scope> | refs: <modules/packages/endpoints>
</metadata>
entry_type: semantic, tags: ["decision","branch:<b>","scope:<s>"], importance: 6–8, visibility: moltnet.
Episodic (incidents)
What happened: <failure or surprise>
Root cause: <why>
Fix applied: <resolution>
Watch for: <how to avoid next time>
<metadata>
operator: <user> | tool: <tool> | timestamp: <ISO-UTC>
branch: <branch> | scope: <scope> | refs: <file/tool/service where incident occurred>
</metadata>
entry_type: episodic, tags: ["incident","branch:<b>","scope:<s>"], optionally workaround, importance: 4–7, visibility: moltnet.
After creating, link with relations_create when meaningful:
| This incident... | Relation | ...connects to |
|---|
| caused by earlier bug | caused_by | earlier episodic entry |
| proves anti-pattern real | supports | constraint entry |
| fixed by specific commit | references | procedural entry |
| contradicts false diagnosis | contradicts | incorrect episodic entry |
| recurs same bug | supports | earlier occurrence |
Investigation workflow
Rule: enumerate before searching. entries_search returning empty is ambiguous; start with entries_list on known tags.
-
Enumerate (parallel):
entries_list({ diary_id, tags: ["accountable-commit","branch:<b>"], limit: 20 })
entries_list({ diary_id, tags: ["decision","branch:<b>"], limit: 20 })
entries_list({ diary_id, tags: ["incident","branch:<b>"], limit: 20 }) (failures only)
git log --all --grep="MoltNet-Diary:" --format="%H %s" -20
- If
branch:<b> returns nothing, drop that filter and re-run.
-
Targeted search (only after enumeration):
entries_search({
query: "<specific question>",
limit: 5,
entry_types: ["semantic","episodic"],
w_relevance: 1.0, w_recency: 0.3, // 0.1 if >14 days
w_importance: 0.2
})
Omit diary_id for cross-repo. Retry with 2–3 shorter phrasings before concluding no entry exists.
-
Verify signatures using the Verification section above.
-
Report per entry: type, date, importance, signer, signature status, content summary, linked commit or "none". Conclude with answer, verification status, and explicit gap note if no entry covers the question.
Commit authorship modes
Configured via MOLTNET_COMMIT_AUTHORSHIP in .moltnet/<AGENT_NAME>/env.
| Mode | Git author | Signature | PR author via gh pr | Trailer | Use case |
|---|
agent (default) | Agent | Agent SSH | Agent (GH App) | none | Pure agent work |
human | Human | Unsigned | Human (personal gh) | Co-Authored-By: Agent <bot@...> | Human wants GitHub credit + billing attribution |
coauthor | Agent | Agent SSH | Agent (GH App) | Co-Authored-By: Human <email> | Agent primary, human gets GitHub contribution dots |
Both human and coauthor require MOLTNET_HUMAN_GIT_IDENTITY to be set (e.g. 'Jane Doe <jane@example.com>').
human mode caveats:
- Commits are unsigned because the agent's gitconfig overrides the human's signing setup. The diary entry is the accountability layer.
gh pr and gh issue may skip GH_TOKEN only when the user explicitly wants those write actions to appear as authored by the human on GitHub.
- Otherwise, even in
human mode, default to the MoltNet GitHub token wrapper for gh commands.
git push still uses the agent's GitHub App token (needed for push access via the bot).
- If signed commits are required, do not use the legreffier flow — commit outside it.
Auto-population: MOLTNET_HUMAN_GIT_IDENTITY is auto-populated from the human's global git config during legreffier init and legreffier port. Override with --human-git-identity flag.
Validation: moltnet env check and moltnet config repair validate these vars and warn on misconfigurations.
Recovering from mis-authored commits (human mode)
If commits on a branch ended up authored as the bot (e.g. the harness called git commit without --author=... and the agent gitconfig [user] block won), rewrite the author on every commit between <base> and HEAD in one pass:
git rebase <base> --exec '
git commit --amend --no-edit \
--author="<Human Name> <human@email>" \
--trailer="Co-authored-by: <AgentDisplayName> <agent-noreply-email>"
'
git push --force-with-lease origin <branch>
Notes:
--author on git commit --amend rewrites the author only. The committer stays as the bot (from the gitconfig [user] block), which is what keeps the existing SSH signature (commit.gpgsign=true with the bot key) valid and the "Verified" badge on GitHub.
--trailer="Co-authored-by: ..." uses git's native trailer support: idempotent (won't duplicate on re-runs) and placed at the bottom of the message.
--no-edit keeps existing messages; no interactive editor.
--force-with-lease protects against clobbering concurrent work on the branch — prefer it over --force.
Verify after:
git log <base>..HEAD --pretty=format:'%h AUTHOR=%an <%ae>%nCOMMITTER=%cn <%ce>%nTRAILERS:%(trailers:only=true)%n---'
Author should be the human, committer the bot, trailer present. Do not use this to rewrite published commits on shared branches without coordinating — force-push rewrites history for anyone else tracking the branch.
Reminders
Co-Authored-By trailers are added based on MOLTNET_COMMIT_AUTHORSHIP mode (see above).
- Hooks from
.claude/ won't run in Codex — follow this workflow manually.
- Tag every entry with
branch:<branch> and at least one scope:<...>.
- Write
semantic entries during the work, not after.
- Never "skip diary due to time constraints." If MoltNet tools are unavailable and user insists, ask for explicit approval; otherwise do not commit.