Menu
End-of-session — update memory, write journal entry, clean up
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Post-merge workflow — extract lessons and compile rules (manual sequence, wrap is retired)
Session-start — consume/derive orientation, poll mail since last signoff, re-derive carryforward gates, present next-steps for operator ruling
Drive the local pre-push review loop to settle — absorb findings locally before any external bot pass
Unified PR review triage — fetch, normalize, and batch-action bot comments
Compiled architectural rules and lessons from the Totem knowledge base. Load this when reviewing code for pattern violations or architectural drift.
Pre-work ritual — spec, search, and (when needed) a 1-page implementation design doc before touching code
| name | signoff |
| description | End-of-session — update memory, write journal entry, clean up |
End-of-session wrap-up. Post-Proposal-282 (ADR-106), journals + handoffs live in the per-repo .totem/orchestration/<agent-id>/ tree (gitignored) — NOT the substrate. Substrate stays as a frozen archive for forensic reads; the active surface is local.
Update memory. Update auto-memory files (e.g. MEMORY.md, topic memories) with any new state — version shipped, tickets closed, key decisions, banked feedback or doctrine signals.
Write a journal entry to the per-repo orchestration path. Filename convention: <model>-NNNN-<short-topic-slug>.md (e.g., claude-0057-phase-4-resolver-shipped.md).
Resolve the path two steps:
a. Identify your agent-id from the current repo's basename. The hardcoded map (Proposal 282 § Scope item 3 — keep in sync with the ADR-106 cohort list):
Repo (git rev-parse --show-toplevel basename) | Claude agent-id | Gemini agent-id | Kimi agent-id |
|---|---|---|---|
totem | totem-claude | totem-gemini | totem-kimi |
totem-strategy | strategy-claude | strategy-gemini | (not seated) |
liquid-city | lc-claude | lc-gemini | (not seated) |
arhgap11 | arhgap11-claude | arhgap11-gemini | (not seated) |
totem-status | status-claude | status-gemini | (not seated) |
totem-playground | (orphan stream — no native agent) | (orphan stream) | (orphan stream) |
Seat discovery is dir-derived (mmnto-ai/totem#2141): any .totem/orchestration/<agent-id>/ directory registers that seat for this repo, UNIONED with the basename map above so roster siblings stay visible on fresh clones where the gitignored tree is partial (precedence: TOTEM_SELF_AGENT env > config.json host_agents > seat dirs ∪ basename map). Override hook: a host_agents: string[] field in .totem/orchestration/config.json still replaces the derived answer — but omitting a PRESENT seat dir attaches a loud warning naming the omitted seat (the dir is the registration; config-exclusion is not a decommission mechanism). The returned list of agent-ids is used by consumers (e.g., totem mail) to filter cross-repo handoffs — messages addressed to any agent-id in the list belong to this repo's session.
Visiting case. If your row's Claude-agent-id column is _(no Claude variant)_ or _(orphan stream — no native agent)_, you are visiting a repo that doesn't natively host your agent. Resolve the journal path to <repoRoot>/.totem/orchestration/<your-home-agent-id>/journal/, where <your-home-agent-id> is your own agent-id (e.g., a strategy-claude session always writes as strategy-claude regardless of which repo it's visiting; concretely, strategy-claude visiting totem-status writes to totem-status/.totem/orchestration/strategy-claude/journal/). The journal records the visiting agent's session state — the host repo doesn't need a native Claude agent to be a valid write target.
b. Resolve the journal directory via resolveOrchestrationPaths(repoRoot, agentId).journal from @mmnto/totem. Returns the absolute path to <repoRoot>/.totem/orchestration/<agent-id>/journal/ when the tree exists. If source === 'none' (the tree does not exist yet in this repo) the resolver returns null for every path field — in that case, construct the path manually as <repoRoot>/.totem/orchestration/<agent-id>/journal/ and create the directory first via mkdir -p; the path is gitignored and safe to create.
No commit, no push. .totem/orchestration/ is gitignored — local filesystem write is the entire operation. No more substrate rebase-retry loops; the cross-agent write-collision class is eliminated by the single-writer-per-path invariant (you only ever write into your own <agent-id>/ subtree).
Clean up stale local branches:
git for-each-ref --format='%(refname:short) %(upstream:track)' refs/heads | while read -r branch track; do
[[ "$track" == "[gone]" ]] && git branch -D -- "$branch"
done
Prune + compact your own ECL cursor (retention + processed-mark GC). Delete your own outbox/ dispatches older than the retention window (N = 14 days) per ECL outbox-retention doctrine (mmnto-ai/totem-strategy:doctrine/ecl-discipline.md § 4.4), THEN compact your processed/ cursor per § 4.5 / ADR-106 § A2. The outbox is transport, not archive — a dispatch's durable content already lives in its home (rulings → ADRs / issues, work-state → the GH board, session history → journal/), so the aged courier file is disposable (gitignored + local). The processed/ cursor is the read-side twin: a mark whose inbound dispatch its sender already swept shadows nothing, so it is safely collectable. The operator should never have to janitor the mail substrate.
Mechanism: totem ecl-gc --apply --compact — self-resolves your agent-id (same precedence as step 2a: TOTEM_SELF_AGENT env > config.json host_agents > seat-dir ∪ basename map). It prunes only <repoRoot>/.totem/orchestration/<your-agent-id>/outbox/ (a self-resolving binary structurally cannot prune a peer), then compacts only your own processed/ marks that shadow nothing. Compaction is cursor-coupled, not age-based, and deletes ONLY against a provably-complete poll — full expected cohort roster present, zero scan warnings, not truncated — else it retains everything (uncertain ⇒ retain). Dry-run by default; --apply deletes. Neither phase touches journal/. Report the pruned + collected counts. Exit codes: 0 clean · 1 some deletes failed (janitorial sensor) · 2 usage/agent-unresolvable · 3 compaction ABORTED loudly (fail-loud, never a silent skip) — no cohort roster declared, the roster is incomplete on this machine, or its A2.4 re-poll check tripped. Do not block the seal on 1 or 3 — the gate-red arms retain the whole cursor (uncertain ⇒ retain); only note them. The gc is a janitorial sensor, not a gate (Tenet 13).
Report: what shipped, what's pending, what's next.
Cross-repo handoffs (when you need to dispatch a message to another agent) write to your own <repoRoot>/.totem/orchestration/<agent-id>/outbox/<YYYY-MM-DDTHHMMZ>-<your-agent-id>.md with to: <recipient-agent-id> in the frontmatter. Recipients discover inbound handoffs by polling the single-level glob <workspace>/*/.totem/orchestration/*/outbox/*.md filtered by their own to: frontmatter match.
Substrate (legacy) is read-only. Do NOT write new content to mmnto-ai/totem-substrate:.handoff/ or :.journal/. The substrate stays mounted as a frozen archive accessible via resolveSubstratePaths(cwd) for forensic reads; the cutover broadcast (when it lands) will confirm the final substrate-write cutoff.