checkpoint-log
Skill27 starsChanged 18 days agoApache-2.0View on GitHub
Keep a live, committed checkpoint log during any work unit expected to span multiple commits or sessions. Each commit ships with a short note in the log, what the commit did, the key decision and why, any deviation from plan; closing the unit requires an explicit walk of the checklist including the wiring steps. Triggers when starting multi-commit work, when resuming a unit after a crash or a gap, and when declaring a multi-commit work unit done. Does not fire on single-commit fixes, where the commit message carries the note.
In this repository56 other files it gives its agents.
- ai-surface-disciplineSkill
- blast-radius-grepSkill
- commission-the-rosterSkill
- committee-reviewSkill
- confirm-before-pushSkill
- constant-time-secret-compareSkill
- context-economySkill
- contract-reviewSkill
- cross-agent-reviewSkill
- date-parse-utc-safeSkill
- db-migration-verificationSkill
- deliverable-integritySkill
- dependency-upgrade-verificationSkill
- earn-every-lineSkill
- email-delivery-verificationSkill
- enforce-invariants-in-buildSkill
- env-change-verificationSkill
- findings-are-evidenceSkill
- fixtures-from-realitySkill
- flag-deferred-itemsSkill
- guard-the-spend-pathsSkill
- handoverSkill
- honest-failure-surfacingSkill
- legal-citation-integritySkill
- legal-notice-handlingSkill
- live-data-surgerySkill
- live-state-firstSkill
- lock-at-the-chokepointSkill
- mass-red-triageSkill
- measure-per-routeSkill
- no-silent-data-dropSkill
- one-real-rideSkill
- outbound-side-effect-idempotencySkill
- parallel-work-reconSkill
- pin-to-what-was-seenSkill
- plan-firstSkill
- price-the-spendSkill
- prove-it-can-failSkill
- reproduce-the-real-buildSkill
- rerun-before-verdictSkill
- safe-smokesSkill
- secrets-in-outputSkill
- server-side-authoritySkill
- skill-library-builderSkill
- spawn-from-declared-stateSkill
- statute-to-obligations-registerSkill
- substantiate-outward-claimsSkill
- team-loopSkill
- tool-layer-bytesSkill
- trace-one-recordSkill
- unordered-eventsSkill
- unslop-codeSkill
- unslop-textSkill
- unslop-uiSkill
- verified-citationsSkill
- verify-the-effectSkill
The file
Files
Checkpoint log
A handover note is written at the end of a session, so a session that dies unexpectedly takes its unwritten detail with it, and the next session inherits a compressed summary it has to re-derive. This skill moves the record to the commit boundary: every commit in a multi-commit unit carries a short checkpoint note, so the repository sits at a documented, resumable checkpoint after every commit. Layer: leaf (commit seam). Admitted by review rather than by the misses log — see Provenance — so candidate-tier until fires accumulate.
The rule
A multi-commit work unit keeps a live log entry per commit, staged with that commit, and closes only after an explicit item-by-item walk of its checklist.
How to apply
- The file.
WORKLOG.mdin the project root, append-only, one##heading per work unit. A project with an established worklog location uses that instead, consistently. - The unit opens with two things: a one-line goal, and the checklist of planned commits or tasks as checkboxes. plan-first produces this list; copying it here makes progress visible in the file, not only in chat.
- The per-commit entry is written before the commit and staged with the code: two to four sentences — what the commit did, the key decision and why it went that way, any deviation from the plan, any finding worth keeping. Tick the checklist item in the same edit. Long evidence goes in the commit body or the project's docs, never the log.
- Why inside the commit: any commit can then be checked out cold and the log explains how the work got there; a later session resumes from the last entry without re-deriving, and the closing summary is written from the log alone.
- Closing the unit: walk the checklist explicitly, item by item, before declaring it done — including the wiring steps that are easiest to skip: registrations and DI bindings, route tables and exports, config and environment entries, gitignore entries, migration catalog queries (db-migration-verification), doc updates, and deferred-item anchors planted where promised (flag-deferred-items). Only then write the closing summary from the log: what was built, what deviated, what was deferred and where its anchor is.
What this skill does not do
It does not serve the session seam — handover owns that, and synthesises its landed, in-flight, and deferred sections from this log where one exists. It does not probe repo or live state at boundaries (parallel-work-recon, live-state-first). It does not decide the plan (plan-first) and does not gate "done" on deployed surfaces (one-real-ride owns that moment; the closing walk here is about the checklist being genuinely finished, not the change being genuinely live).
Why
Mid-unit failure is where detail is lost: "why" evaporates faster than "what", and the closing summary written from memory is the one that omits the deviation that mattered. A note per commit keeps why next to what while both are fresh, makes every commit a safe stopping point, and turns the closing summary into a synthesis job rather than a memory job. The closing walk exists because the items that break a "finished" unit are rarely the interesting code; they are the wiring steps nobody remembers doing.
Provenance
Adapted from the CURRENT_PHASE checkpoint pattern and phase-close checklist rule in Project Architect 2.0 (source-available; reviewed 2026-08-03, verdict: do not adopt wholesale, nick this pattern). Text written in-house; no upstream text copied.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

