agentleFS
jackneiljackneil/claude-jacked/.claude/skills/release/SKILL.md

release

Skill2 starsChanged 11 days agoMITView on GitHub

Cut a release the way this repo actually ships: detects PyPI/npm/changesets/CalVer/Cargo/Go-tag/PR-to-main, gates on the repo's own build+test, then publishes/deploys and verifies it landed. (repo)

  • Deletes or force-pushes
  • Installs packages
  • Commits and pushes

193 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use create-agent-adapter, changelog-style and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • MicrosoftSkill

    create-agent-adapter

    Technical guide for creating a new Paperclip agent adapter. Use when building a new adapter package, adding support for a new AI coding tool (e.g. a new CLI agent, API-based agent, or custom process), or when modifying the adapter system. Covers the required interfaces, module structure, registration points, and conventions derived from the existing claude-local and codex-local adapters.

    Also in 5 projects that use this

  • GrinvSkill

    changelog-style

    Turn a commit range into CHANGELOG.md entries following common-changelog style. Use when finishing a PR or cutting a release, to write/fix the [Unreleased] section's entries — moving [Unreleased] under a new dated version heading is a separate, automated step (see the release skill / scripts/sync-version.ts), not this skill's job.

    Also in 4 projects that use this

  • GrinvSkill

    docs-consistency-check

    Check README/manifest.json/server.json/CHANGELOG.md/AGENTS.md and docs/*.md for drift against the actual registered tools and source. Use after adding, renaming, or removing a tool, or as part of a live-audit pass.

    Also in 4 projects that use this

  • Microsoft AzureSkill

    component-release

    Prepares a verifiable Agent Landing Zone UI release. Use for release branches, VERSION, changelog entries, tags, release notes, and compatibility evidence; never publish without explicit human approval.

    Also in 4 projects that use this

  • Microsoft AzureSkill

    documentation-consistency

    Keeps AI Landing Zone contributor, user, and operator documentation aligned with shipped Bicep behavior. Use for parameters, defaults, outputs, topology, deployment flow, operations, or releases.

    Also in 8 projects that use this

  • rivet-devSkill

    launch-graphics

    Generate agentOS marketing/launch graphics (launch & social heroes, code-snippet cards) from the committed generators in docs-internal/graphics. Use when asked to make, update, or render an agentOS launch image, social image, "agentOS Apps" graphic, or technical code-snippet image.

    Also in 3 projects that use this

  • +3 more from Grinv

The file

Show its rules and text

We split this skill into the rules it gives an agent

18 rules. 0 are common, 8 are distinctive, and 10 are very specific to this project.

  • 0

    Common

    Most of the 183 edited versions we read keep these.

  • 8

    Distinctive

    This author's own choices. Worth a look.

  • 10

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Before using the config above, run a lightweight drift check (advisory only.

    line 46
  2. Very specific:

    If EITHER differs, count the intervening commits (git rev-list --count <generated-commit>..HEAD 2>/dev/null) and surface one line, then …

    line 49
  3. Distinctive:

    If the generated commit is unreachable (history rewritten) or the manifest set changed, surface the nudge regardless of commit count.

    line 51
  4. Very specific:

    If invoked via a local config wrapper (you see a ## Repo Config section earlier in the prompt), it is authoritative — use it, do not …

    line 60
  5. Very specific:

    If no ## Repo Config is present, auto-detect the model in the next section.

    line 68
  6. Distinctive:

    Run in order. Stop and report if any step fails.

    line 102
  7. Very specific:

    Only when the gate is fully green do you proceed.

    line 146
  8. Very specific:

    If a CHANGELOG.md exists (or the README carries a "Version History" section), prepend the new version's delta (grouped …

    line 150
  9. Distinctive:

    If push fails on upstream changes, stop and ask (never force-push without explicit approval).

    line 165
  10. Very specific:

    pypi: create the GitHub Release — this triggers the publish workflow (gh release create vX.Y.Z --title "vX.Y.Z" --generate-notes --latest).

    line 169
  11. Very specific:

    npm / calver: if a workflow publishes on Release/tag, cut the Release and watch that run as above.

    line 170
  12. Very specific:

    If a publish/deploy run fails: gh run view <id> --log-failed, report it, and never leave a --latest Release (or a "shipped" claim) pointing …

    line 177
  13. Very specific:

    After the release is verified live, if the memory vault is enabled (jacked memory status --quiet exits 0.

    line 198
  14. Distinctive:

    Read the shipping model from config or detect it FIRST.

    line 208
  15. Distinctive:

    Never publish to a package index for a repo that has no package (pr-to-main-deploy / private app).

    line 209
  16. Very specific:

    Never force-push without explicit approval.

    line 211
  17. Distinctive:

    Never skip the version bump for a versioned model (indexes reject duplicate versions) — and never bump for a model that doesn't version …

    line 212
  18. Distinctive:

    If anything fails, stop and report — do not retry destructively.

    line 214

“Common” is measured: kept by at least 6 in 10 of the 183 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Generated by /jacked-setup - 2026-11-03 | Template v2 | Engine: jacked v0.110.2

Standalone - no dependencies. Commit this file. To update: uv tool install --upgrade claude-jacked && jacked install && /jacked-setup release

Repo Config

  • Project: claude-jacked
  • Base Branch: master
  • Shipping Model: pypi
  • Version Source: jacked/__init__.py __version__ (hatch dynamic version: pyproject.toml declares dynamic = ["version"] with [tool.hatch.version] path = "jacked/__init__.py"; do NOT add a static version to pyproject.toml)
  • Gate Commands: mirror .github/workflows/tests.yml, then the release build of .github/workflows/build-dist.yml (the job publish.yml runs; --no-config keeps the project exclude-newer out of a pip-equivalent install). Run them from the repo root in this order, each as cmd > log 2>&1; echo EXIT:$?, and quote pytest's own summary line; never trust a piped exit code. CI runs pytest on ubuntu, macos and windows; the ruff rule set is pinned in [tool.ruff.lint]. The two audit tools come from the hash-locked .github/requirements/ci-audit.txt, not from the dev group.
    uv lock --check
    uv run --frozen ruff check .
    uv run --frozen python -m pytest
    GATE_TMP="$(mktemp -d)"
    uv export --frozen --format requirements-txt --no-emit-project --no-hashes --no-annotate --no-header -o "$GATE_TMP/export.txt"
    sed -E 's/ *;.*$//' "$GATE_TMP/export.txt" | sort -u > "$GATE_TMP/locked.txt"
    uv run --frozen python - "$GATE_TMP/locked.txt" <<'PY'
    import re, sys
    lock = open("uv.lock", encoding="utf-8").read()
    locked = {f"{n}=={v}".lower() for n, v in re.findall(r'\[\[package\]\]\nname = "([^"]+)"\nversion = "([^"]+)"\nsource = \{ registry', lock)}
    audited = {line.strip().lower() for line in open(sys.argv[1], encoding="utf-8") if line.strip()}
    print(f"auditing {len(audited)} of {len(locked)} locked packages")
    sys.exit(0 if locked and audited == locked else f"coverage mismatch: missing {sorted(locked - audited)}, extra {sorted(audited - locked)}")
    PY
    uv run --isolated --no-project --with-requirements .github/requirements/ci-audit.txt pip-audit -r "$GATE_TMP/locked.txt" --no-deps --disable-pip --vulnerability-service osv
    uv run --isolated --no-project --with-requirements .github/requirements/ci-audit.txt zizmor --no-progress --strict-collection .github/workflows
    uv venv -q -p 3.11 "$GATE_TMP/pub"
    uv pip install -q --no-config -p "$GATE_TMP/pub" --require-hashes --only-binary=:all: -r .github/requirements/ci-publish.txt
    uv run --no-project --no-config -p "$GATE_TMP/pub" python -m build --no-isolation --outdir "$GATE_TMP/dist"
    uv run --no-project --no-config -p "$GATE_TMP/pub" twine check "$GATE_TMP"/dist/*
    uv run --no-project --no-config -p "$GATE_TMP/pub" cyclonedx-py environment -o "$GATE_TMP/sbom.cdx.json"
  • Publish Mechanism: GitHub Release published on master -> .github/workflows/publish.yml (build job = .github/workflows/build-dist.yml: hash-locked python -m build --no-isolation, twine check dist/*, CycloneDX SBOM; publish job in the pypi environment) -> PyPI Trusted Publisher (OIDC, pypa/gh-action-pypi-publish) -> https://pypi.org/p/claude-jacked. Cutting the GitHub Release (e.g. gh release create vX.Y.Z) IS the publish trigger: treat it as the irreversible step.
  • Verify: poll https://pypi.org/pypi/claude-jacked/json until the new version appears; confirm the publish.yml run succeeded (gh run list --workflow=publish.yml); optionally uv tool install --upgrade claude-jacked and check jacked --version. Before tagging, confirm README's "Version History" table carries the new version's row, and that the merged PR's Tests run shows a real "N passed" line on each OS leg (a green check alone has hidden a pytest that died early with exit 0).
<!-- Staleness fingerprint - the engine compares these against the live repo on each run -->
  • Generated at commit: b3c14640091e7653e438cfcb1c2254f610c29550
  • Detected stack: pyproject.toml

Config Staleness Check

Before using the config above, run a lightweight drift check (advisory only; never block on it):

  1. git rev-parse HEAD and compare to Generated at commit above.
  2. ls package.json pyproject.toml go.mod Cargo.toml setup.py Gemfile pom.xml build.gradle composer.json mix.exs 2>/dev/null and compare the manifest set to Detected stack above.
  3. If EITHER differs, count the intervening commits (git rev-list --count <generated-commit>..HEAD 2>/dev/null) and surface one line, then continue:

    ⚠️ This /release config was generated N commits ago and the stack/paths may have drifted. Re-run /jacked-setup release to refresh. If the generated commit is unreachable (history rewritten) or the manifest set changed, surface the nudge regardless of commit count. Never stop or skip the run on a staleness mismatch.

<!-- ENGINE - DO NOT EDIT BELOW THIS LINE -->

You are the Release Manager. Cut a release for this repo — however it actually ships. Do NOT assume it is a Python/PyPI package: read the shipping model from config (or detect it) FIRST, then run that model's pipeline.

Config Override

If invoked via a local config wrapper (you see a ## Repo Config section earlier in the prompt), it is authoritative — use it, do not re-detect:

  • Base Branch → the branch to release from / push to (not a hardcoded master).
  • Shipping Model → which pipeline below to run (pypi / npm / npm-changesets / calver / cargo / go-module-tag / pr-to-main-deploy / github-release-only).
  • Version Source → the exact file + field to read and bump (or none when nothing is versioned).
  • Gate Commands → the repo's real build/test/lint to run at Step 4 instead of uv build/twine/pytest.
  • Publish Mechanism → how it actually ships (a GitHub Release that triggers a workflow, npm publish, changeset publish, a merge/PR to a deploy branch, a bare tag, …).
  • Verify → how to confirm it landed (poll the package index, or confirm the deploy run succeeded + the app is up).

If no ## Repo Config is present, auto-detect the model in the next section.

SAFETY — this command does irreversible things (publishes, tags, deploys):

  • The config/detected model decides the publish target. If the model or target is ambiguous, or you detect more than one plausible target, STOP and ask — never guess where to publish or which branch to push.
  • Never publish to a package index for a repo that has no package (a private app / pr-to-main-deploy repo). Never invent a PyPI/npm step that the repo's own CI doesn't already do.
  • Everything reversible (version bump, gate) happens BEFORE anything irreversible (tag, publish, merge-to-deploy-branch). Never force-push without explicit approval.

Auto-detect the shipping model (when no Repo Config)

Detect conservatively — read, don't assume:

ls pyproject.toml setup.py package.json Cargo.toml go.mod 2>/dev/null   # ecosystem
ls -d .changeset 2>/dev/null                                            # npm changesets
ls .github/workflows/ 2>/dev/null
# What does CI actually DO on release/push/tag? (the source of truth for how it ships)
grep -rlE 'twine|pypi|gh-action-pypi|npm publish|changeset|cargo publish|railway|vercel|fly\.io|flyctl|deploy' .github/workflows/ 2>/dev/null
grep -rlE 'on:|release:|push:|tags:|workflow_dispatch:' .github/workflows/ 2>/dev/null
git remote -v 2>/dev/null; git symbolic-ref refs/remotes/origin/HEAD 2>/dev/null | sed 's|refs/remotes/origin/||'

Map signals → model (priority order; when two genuinely fit, ask):

  • .changeset/ + package.json → npm-changesets (changeset version bumps, changeset publish ships).
  • package.json + an npm publish / publishConfig in a workflow → npm — but if versions are date-shaped (2026.318.0) or a workflow derives the version from the date → calver.
  • pyproject.toml/setup.py + a workflow that publishes to PyPI (twine / OIDC trusted publisher) → pypi.
  • Cargo.toml + cargo publish → cargo.
  • go.mod, no publish step → go-module-tag (the tag itself IS the release; the module proxy serves it).
  • A workflow that deploys on push/merge to the base branch (railway/vercel/fly/a deploy step) with NO package publish → pr-to-main-deploy: "shipping" = getting the change onto the deploy branch; there is no version bump or index publish, and cutting a package would be wrong.
  • None of the above, but you still want a GitHub Release for notes → github-release-only (tag + Release, no publish).

State the detected model and the evidence you based it on. If unsure, ask before touching anything.

RELEASE PIPELINE

Run in order. Stop and report if any step fails.

1. PRE-FLIGHT

BASE="${BASE_BRANCH:-$(git symbolic-ref refs/remotes/origin/HEAD 2>/dev/null | sed 's|refs/remotes/origin/||' || echo main)}"
git branch --show-current      # should be the base branch (or the branch you're releasing)
git status --short             # working tree state
gh auth status                 # GitHub auth (needed for Releases / PR merges)
  • Not on the base branch → stop; ask whether to merge first or release from this branch.
  • Uncommitted changes → list them; ask whether to include them in the release or stash.
  • gh not authenticated → stop; tell the user to run gh auth login.

2. DETERMINE VERSION

Skip this step entirely for calver (the version is computed, usually date-based — read how from the workflow) and pr-to-main-deploy / any none Version Source (nothing is versioned — the merge is the ship).

Otherwise read the current version from the Version Source:

  • Config declares it → use that file+field.
  • Else detect: pyproject.toml [project].version if static; if dynamic = ["version"], follow [tool.hatch.version].path (or grep -rl __version__) to the real file; package.json version; Cargo.toml [package].version. In a monorepo, the versioned artifact may live in a workspace member, not the root — read the member that actually ships.

Suggest the bump from commit history (the user still picks):

git describe --tags --abbrev=0 2>/dev/null
git log "$(git describe --tags --abbrev=0 2>/dev/null)"..HEAD --oneline 2>/dev/null

BREAKING CHANGE/type!: → major; any feat: → minor; only fix:/chore:/docs: → patch. If $ARGUMENTS has a version, use it. Otherwise ask, leading with the suggestion (Patch / Minor / Major / Custom).

3. BUMP VERSION

Edit the Version Source to the new version (__version__, [project].version, package.json "version", Cargo.toml, etc.). For npm-changesets, run the changeset flow instead (pnpm changeset → changeset version). Do NOT commit or tag yet — the gate (Step 4) must pass first. Skip for calver/none.

4. GATE — before anything irreversible

The tag/Release/publish/deploy are what ship the code, and many repos' CI does NOT run tests on release — so catch a broken build or red test HERE, while everything is still local and reversible. Run the repo's own gate:

  • Config Gate Commands → run exactly those, in order.
  • Else per model (run what the repo actually uses — read package.json scripts / the CI workflow, don't assume):
    • pypi: rm -rf dist/ && uv build && uvx twine check dist/* && uv run python -m pytest
    • npm / npm-changesets / calver: the repo's real chain, e.g. pnpm -r typecheck && pnpm -r test && pnpm -r build (or npm ci && npm test && npm run build; turbo repos: turbo run typecheck test build).
    • cargo: cargo build --release && cargo test.
    • go-module-tag: go build ./... && go test ./....
    • pr-to-main-deploy: run the repo's CI gate locally (the lint/typecheck/test/build the deploy workflow runs) so the merge doesn't break the deploy.

Only when the gate is fully green do you proceed. If it fails, fix the cause and re-run — do NOT tag/publish/merge a red tree.

5. CHANGELOG + COMMIT + TAG

If a CHANGELOG.md exists (or the README carries a "Version History" section), prepend the new version's delta (grouped Features/Fixes/Other) before committing, so the tagged tree carries its own changelog. Then:

git add <version file> <changelog>   # + any other approved changes
git commit -m "chore: release vX.Y.Z"   # or feat/fix: vX.Y.Z — <summary> for substantive changes
git tag vX.Y.Z                          # SKIP the tag for pr-to-main-deploy (the merge is the ship; tag only if the repo tags)

6. PUSH / SHIP

  • pypi / npm / cargo / go-module-tag / github-release-only: push the base branch + tag.
    git push origin "$BASE" && git push origin vX.Y.Z
  • pr-to-main-deploy: getting the change onto the deploy branch IS the release. If you're on a feature branch, open a PR to the deploy branch and (once its CI is green) merge with a true merge commit (gh pr merge --merge, never --squash/--rebase/--admin); if you're already on the deploy branch with approval, push it. The merge/push triggers the deploy — there is no package to publish.

If push fails on upstream changes, stop and ask (never force-push without explicit approval).

7. PUBLISH (per model)

  • pypi: create the GitHub Release — this triggers the publish workflow (gh release create vX.Y.Z --title "vX.Y.Z" --generate-notes --latest). Then watch it: find the run for THIS tag (gh run list --workflow=<publish>.yml and select the one whose headBranch == vX.Y.Z — do not grab a stale prior run) and gh run watch <id> --exit-status.
  • npm / calver: if a workflow publishes on Release/tag, cut the Release and watch that run as above. If publishing is manual, run the repo's publish (npm publish / pnpm -r publish) with the right registry/OTP.
  • npm-changesets: pnpm changeset publish (or let the Changesets Action open/merge the version PR, then publish) — follow the repo's setup; watch the action.
  • cargo: cargo publish (per crate, in dependency order for a workspace).
  • go-module-tag: nothing to publish — the pushed tag is the release. Optionally gh release create for notes.
  • github-release-only: gh release create vX.Y.Z --generate-notes --latest; no index publish.
  • pr-to-main-deploy: nothing to publish. Publishing here = watching the DEPLOY: find the deploy workflow run triggered by the merge and gh run watch <id> --exit-status; a red deploy is a failed release.

If a publish/deploy run fails: gh run view <id> --log-failed, report it, and never leave a --latest Release (or a "shipped" claim) pointing at a version/deploy that isn't actually live — re-run from Actions (keeps the tag) or delete the Release+tag and re-cut after a fix.

8. VERIFY IT LANDED

A green workflow means "the job succeeded," not "it's live." Confirm reality:

  • pypi: poll the index until the version resolves, then prove installability (derive the name from pyproject.toml, don't hardcode):
    NAME=$(grep -m1 '^name' pyproject.toml | sed -E 's/.*"([^"]+)".*/\1/')
    curl -fsSL "https://pypi.org/pypi/${NAME}/json" | grep -q '"X.Y.Z"'   # JSON API
    # and the /simple/ index (what installers resolve) — it can lag the JSON API by a minute
    # then, from a NEUTRAL dir (outside the repo, to avoid a pinned exclude-newer): uv tool install "${NAME}@X.Y.Z" --force --refresh
  • npm / calver: npm view <pkg>@X.Y.Z version until it resolves (and npm install in a throwaway dir for the strong proof).
  • cargo: cargo search <crate> / crates.io shows the version.
  • go-module-tag: go list -m <module>@vX.Y.Z resolves via the proxy.
  • pr-to-main-deploy: the deploy run is green AND the app is actually up — curl -fsSItimeout the production/preview URL (from config or the deploy output) for a 2xx, and sanity-check a health route if one exists. A green deploy job with a 502 site is a failed release.

Then report the concrete artifact: the Release URL, the index/registry URL + install line, or the live app URL — whichever the model produced.

9. RECORD (memory vault, guarded)

After the release is verified live, if the memory vault is enabled (jacked memory status --quiet exits 0; skip silently if it exits nonzero), record a progress note so the shipped version and its highlights are searchable later:

jacked memory add --type progress --title "Released vX.Y.Z" --body "<the headline changes + the artifact URL>"

One note per release, high-signal. If the vault is off, do nothing.

HARD RULES

  • Read the shipping model from config or detect it FIRST; when the model or publish target is ambiguous, STOP and ask. Never guess a publish target or push to the wrong branch.
  • Never publish to a package index for a repo that has no package (pr-to-main-deploy / private app). Never invent a publish step the repo's own CI doesn't do.
  • Reversible before irreversible: gate (build + tests, the repo's OWN) must be green BEFORE any tag/publish/merge-to-deploy. Never tag/ship a red tree.
  • Never force-push without explicit approval. Never --squash/--rebase/--admin-merge a release PR — use a true merge commit.
  • Never skip the version bump for a versioned model (indexes reject duplicate versions) — and never bump for a model that doesn't version (calver/pr-to-main-deploy).
  • A tag/Release must NOT outlive a failed publish/deploy — verify the artifact is actually live/installable (or the deploy is actually up) before declaring done.
  • If anything fails, stop and report — do not retry destructively.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.