release-cut
Skill2 starsChanged 52 days agoMITView on GitHub
Cut and verify a CADENCE release end-to-end — inventory unreleased phases, audit changesets, lockstep version bump, a full doc-sync verification pass (automated doc tests + manual stale-version-reference sweep), release PR, operator-triggered Release workflow, and independent npm/tag/GitHub verification. Use when the user says "cut a release", "release vX.Y", "publish to npm", or when a release phase begins.
- Commits and pushes
3 other projects use a version of this.
Adopted by
Most recently taken up by
- 4gray/iptvnatortheir own version, first seen Nov 1, 2026
- tony/skillstheir own version, first seen Oct 31, 2026
- Megaprompting/torque-looptheir own version, first seen Jul 3, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- 4gray/iptvnator★ 7.8ksame name
- Megaprompting/torque-loop★ 10same name
- tony/skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository5 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- phase-buildSkill
- pr-landSkill
- systematic-debuggingSkill
The file
We split this skill into the rules it gives an agent
10 rules. 0 are common, 2 are distinctive, and 8 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
8
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 18
Audit changesets: every feature PR in the bundle should have left a .changeset/*.md.
- Very specific: line 26
Verify all five published packages (core, types, host-claude-code, host-codex, host-toolkit) moved to the identical new version and testkit …
- Very specific: line 29
Update the version line near the top of CLAUDE.md to the new version string.
- Very specific: line 32
Add a matching ## [x.y.z] heading to CHANGELOG.md for the new version, in the same commit.
- Very specific: line 37
Run the full pipeline locally: pnpm turbo run lint typecheck test build.
- Very specific: line 46
Run the doc-content test surface explicitly and confirm it's green: pnpm --filter @thomas-powers-jr/cadence-core test -- tests/docs, pnpm …
- Very specific: line 76
After the release PR merges, ask for the explicit go-ahead, then trigger the manual Release workflow (gh workflow run Release / …
- Distinctive: line 88
Run every check below, regardless of what the workflow says:
- Very specific: line 101
All five on npm + tag + release page present → the run's red was cosmetic.
- Distinctive: line 117
Write the session handoff: version live, bundle list, any slips or flakes encountered, next candidates.
“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Release cut
Publishing is irreversible. Two hard consent gates in this flow: merging the release PR, and firing the Release workflow. Neither happens on a generic "continue" — restate the action and get an explicit go-ahead.
1 — Inventory what ships
git fetch origin --prune, thengit log $(git describe --tags --abbrev=0)..origin/main --oneline --no-merges.- Map commits to phases; write the bundle list (this becomes the PR body and the CLAUDE.md version-line summary).
- Audit changesets: every feature PR in the bundle should have left a
.changeset/*.md. For any missing one (discipline has slipped before — five had to be reconstructed retroactively for v1.42.0), reconstruct it from the PR description now, and note the slip in the handoff.
2 — Version bump (lockstep)
- On a release branch:
pnpm changeset:version. - Verify all five published packages (
core,types,host-claude-code,host-codex,host-toolkit) moved to the identical new version andtestkitdid not. Anything else is a stop-and-investigate. - Update the version line near the top of
CLAUDE.mdto the new version string. The doc-sync pre-commit hook aborts the commit otherwise, andpackages/core/tests/docs/doc-sync-hook.test.tsre-asserts it in CI. - Add a matching
## [x.y.z]heading toCHANGELOG.mdfor the new version, in the same commit. The same doc-sync pre-commit hook now checks CHANGELOG.md too — it aborts the commit if the newest heading doesn't match — andpackages/core/tests/docs/doc-sync-hook.test.tshas a seconddescribeblock that re-asserts it in CI. - Run the full pipeline locally:
pnpm turbo run lint typecheck test build.
3 — Doc-sync verification (mandatory, not implied by the pipeline run)
Step 2's full pipeline only proves the automated doc-content tests pass —
it does not prove every doc mentioning the old version got updated, since
several repo docs (DESIGN.md, docs/*) carry version references with no
test covering them.
- Run the doc-content test surface explicitly and confirm it's green:
pnpm --filter @thomas-powers-jr/cadence-core test -- tests/docs,pnpm --filter @thomas-powers-jr/cadence-host-claude-code test -- docs-command-count docs-published,pnpm --filter @thomas-powers-jr/cadence-host-codex test -- docs-published. - Grep the whole repo for the previous version string and triage every
hit:
grep -rn "<old-version>" --include="*.md" . | grep -v node_modules | grep -v CHANGELOG | grep -v '\.cadence/' | grep -v '\.changeset/'. A hit in a CHANGELOG or a "reconstructed for vX.Y.Z" narrative note is historical and correct as-is; a hit describing current state (e.g.DESIGN.md's "Current architecture (as of vX.Y.Z)" line, which slipped once already in the v1.43.0 cut) needs bumping to the new version. - Confirm every phase in the release has both its
### Phase Nheading in.cadence/ROADMAP.mdand its- **Phase N**bullet in.cadence/MILESTONES.md, and that none still says(in progress). Do not rely oncadence doctor'sroadmap-currencycheck for this: it compares only each file's newest entry against the newest phase directory and warns past a 10-phase drift, so it cannot see a missing entry below either file's newest one. - Note anything fixed (or anything left alone with its reason) in the release PR body and the handoff.
4 — Release PR
- Subject:
chore(release): vX.Y.Z -- <one-line bundle summary>. - Land it via the
pr-landskill (protected main;ci-success,security-successandcodeql-successall required; merge only on explicit operator consent).
5 — Publish (operator-triggered, never automatic)
- After the release PR merges, ask for the explicit go-ahead, then trigger
the manual
Releaseworkflow (gh workflow run Release/.github/workflows/release.yml). - Never
gh run rerun --failedon the Release workflow. It re-runspnpm -r publishand fails on already-published versions. The release-integrity step now waits out registry propagation itself (about 10 minutes of polling, one progress line per miss in the run log), so a red release-integrity step deserves investigation; verify what actually landed by hand (step 6) before touching anything.
6 — Verify independently (never trust the workflow's own report)
Run every check below, regardless of what the workflow says:
npm view @thomas-powers-jr/cadence-core version
npm view @thomas-powers-jr/cadence-types version
npm view @thomas-powers-jr/cadence-host-claude-code version
npm view @thomas-powers-jr/cadence-host-codex version
npm view @thomas-powers-jr/cadence-host-toolkit version
git ls-remote --tags origin | grep vX.Y.Z
gh release view vX.Y.ZDecision table on a red/ambiguous run: check what is actually missing.
All five on npm + tag + release page present → the run's red was cosmetic;
done. Packages on npm but tag/release missing → create only the missing
artifact by hand (git tag -a + push, gh release create). Some packages
missing on npm → if the release-integrity step ran its full poll (its log
shows attempt 40/40), propagation is already over: treat the gap as real and
read the publish step's log for a failed pnpm -r publish before a targeted
republish. If the run stopped before that step, wait out the five-minute
registry edge cache, re-check, and only then consider a targeted republish.
7 — Close the loop
- Settle the release phase (single-commit convention) if it ran as one.
- Promote recommendations now shipped immediately, don't defer it:
settle run --ship-ref "<PR/vX.Y.Z>"in the same settle, orcadence recommendation promote <id> --status=shipped --ref "..."right after publish verification — in the same commit/push, not a later pass. - Write the session handoff: version live, bundle list, any slips or flakes encountered, next candidates.
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.

