agentleFS
KealmanKealman/cland-v0/.agents/skills/release/SKILL.md

release

Skill1 starsChanged 16 days agoApache-2.0View on GitHub

Release a human-chosen version, publish marketing commits outside a release, or recover pending publication.

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

  • 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

  • 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

    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

  • parcadeiSkill

    agent-context-isolation

    Agent Context Isolation

    Also in 3 projects that use this

  • +3 more from Grinv

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 0 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 183 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use cland release --help (or -h) for usage without a hub or Git/network work.

    line 8
  2. Very specific:

    Use release: cland release --item T-n --dry-run.

    line 101
  3. Very specific:

    Use release: cland release --item T-n --dry-run, then ask for notes/version approval.

    line 105

“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

The Coordinator performs this skill from its own shell. Never delegate a release to a sub-agent: a sub-agent in the Coordinator's shell inherits its authority. The human may also run cland release from their own shell.

Use cland release --help (or -h) for usage without a hub or Git/network work; sync, publish, push and merge accept the same help flags. Release, sync and publish share the release usage block.

A release-version item updates the root package version and both lockfile version fields together, and proves --version in a sealed disposable copy under rule 15 (A-8634 R15; T-967).

For cland verify all --occasion release, receipt reuse requires full matching coverage on the effective executor and OS, even with a configured default. A local macOS stage receipt cannot satisfy a droplet release request. Read the cache notice's original occasion/executor; --fresh forces execution. See hub/README.md, Verification commands, for the shared identity contract.

For committed [release] web_bundle = true (cland product only), the release machine needs installed, already-authenticated gh and network access for locked build dependencies. Dry-run checks gh before approval, builds/caches the pinned commit and names repo, tag, archive/checksum filenames, byte sizes and SHA-256. Include all those facts in step 3’s single approval. Generated/older configs default to false. Publication creates a draft as needed, uploads and reads back both assets, then publishes it. On web-bundle follow-up failure, preserve release.tmp and .cland/release-web/, repair the obstruction, and retry the same item/version. Never clobber conflicting remote assets or start a new login. See Web release assets.

For a configured [release.homebrew] tap, include the preview's tap repository, main, Formula/cland.rb, version and exact tag archive URL in step 3's approval. The preview states SHA-256 follows download after the product push. That one approval covers the formula for the approved tag URL; no second approval follows hashing. An empty tap reports skipped. The human creates the tap first. See Homebrew publication for receipt/retry semantics; final output and item comment include SHA and tap commit.

For approval before readiness work finishes, use Conditional release approval before step 2's final dry run. Complete all preflight steps before publication.

  1. Work in the project root on dev. Read the current release item with cland task view T-n --json and use or open its bound topic. The current release is the open top-level item with the lowest semver version. Commit outstanding tracker and docs changes on dev, keeping unrelated staged work out of that commit, and publish them to the territory with cland push from your own shell. Resolve any fast-forward refusal before releasing. The default branch is main unless the territory records another in cland.defaultBranch. Release includes all committed work between the default branch and dev; a project version bump is an ordinary item reviewed and merged before release.
  2. Read the item's version and the human's recorded version decision. When a recorded decision exists, ensure the item's version matches it before previewing. If the version is missing, ask the human to choose it, then record it with cland task edit T-n --version X.Y.Z. Before the dry run, ensure the root package.json version equals the release's through an ordinary commit reviewed and merged onto dev. Release reads the committed dev version and refuses a mismatch with release-version; it never writes the bump. Projects without that file or its version key release as before. Before the dry run, when releasing cland itself (the repository containing hub/bin/check-opencode.ts), run the informational OpenCode check and include any differences in the release report. Differences do not stop the release. Run cland release --item T-n --dry-run. Resolve every refusal before continuing: close or move open descendants; when GitHub has commits dev lacks, run cland sync from a clean root, then preview again. Share the resulting notes, version, commit and refs in the release topic through artifact.create.
  3. For daytime approval while readiness is pending, use Conditional release approval. Otherwise ask through the Coordinator's ask in that topic: show the preview's notes and version and request approval to release that build, allowing a free answer (free: true). Wait for approval. The version comes from the human. An Aide or a stand-in named by the human may relay approval and the version only when the human has recorded that version decision; cite the record (for this project's v0.1, D18 in docs/spec/v0.1.md records v0.1.0). A stand-in never invents a version; an unrecorded version needs the human's decision. Follow the project's recorded acceptance policy: D19 uses QA's clear T-117 walkthrough instead of a separate human acceptance pass, and the notes retain T-109 C1 and T-111 M7 as verified in v0.1.1.
  4. If the approved version differs, record it on the item, run cland release --item T-n --dry-run again and show the resulting notes and version in a fresh approval ask. For ordinary exact-build approval, if dev or the notes changed, preview and obtain approval for that build again. Conditional approval follows its recorded future-change scope below. Once the preview is approved and every recorded condition is satisfied, run cland release --item T-n yourself. It prepares an annotated tag with the item Notes, epic deliverables and commit notes, pushes the default branch and the tag atomically, advances the default branch locally and in the territory, then comments the tag and full commit on the item and closes it. The next open release becomes current. Web assets use gh’s GitHub Release API only when opted in; the Git refs transaction makes no API calls.
  5. Report the returned version and full commit in the release topic. If it says released; follow-up failed, GitHub already accepted the release: preserve .cland/release.tmp, repair the named local, territory, Web assets, Homebrew tap or item obstruction, and retry the same item and version with cland release --item T-n. The receipt pins the published item, tag, commit and notes, so a restart or later dev commit cannot publish a different release on that retry. A failed push can also have lost its reply; retry the same item and version to let cland check the receipt against GitHub. With a pending sync, preserve .cland/sync.tmp and finish cland sync before releasing.

For a project without a release item, cland release <chosen-version> [--dry-run] retains the version-based flow: run the same applicable preflight check, preview the notes, ask for approval with a human-chosen version, then publish from your own shell. It does not comment on or close tracker items.

cland sync fetches GitHub and prepares a replay of dev's unreleased commits. Finish any pending cland merge follow-up first: sync refuses to rewrite a merge whose checkout, item, claim or topic completion is still pending. A dirty root refuses; a conflict names the files and leaves the default branch, dev and the root untouched. Reconcile the conflicting work on dev, then retry. Sync publishes only to the local territory, never to GitHub. If it says synced locally; follow-up failed, preserve its receipt, repair the named obstruction, and retry sync; it finishes the same prepared replay.

Conditional release approval

In daytime, the Coordinator may use ask in the release topic (free: true) to request the human's approval of a named version, notes and explicit conditions before readiness is complete. The ask must name the scope of future changes it authorizes, such as reviewed fixes for named release items and their committed tracker and evidence updates. For example: “Approve release vX.Y.Z with these notes, including reviewed fixes for the listed release items and their tracker and evidence updates, once the release epic is clear and the covers are green?” Attach the notes artifact and identify the release item, current candidate commit, named future work and required covers (platforms, executor and occasions). Name any other conditions in the ask; the human decides both the version and conditions. A declined ask is not approval.

Record the ask id, human answer, approved version, notes artifact and exact conditions and authorized future-change scope in a release item comment. A conditional approval authorizes release only when all those conditions hold. While a dry run still refuses pending readiness, keep the recorded notes as a proposal; complete step 2's successful preview before publication. Compare every intervening commit with the recorded scope. The final successful preview must identify the final head, and required cover receipts must match that head under the existing coverage rules. Existing release gates and acceptance policy apply.

The Coordinator checks each condition and records its evidence in the release item before running cland release: cite the clear epic verdict and matching green cover receipts, including their heads, platforms, executor and occasions, plus evidence for every other named condition. A missing, failed or pending condition means wait; report which evidence is outstanding. When all conditions hold and every intervening commit is within the human-approved scope, release without another ask. Any change outside that scope requires a fresh approval ask. Any change to the version or the notes, conditions or other approved publication facts (including asset and Homebrew facts above) also requires a fresh approval ask. Approval never waives a release refusal.

Publish marketing without a release

cland publish is the only out-of-release GitHub push. The Coordinator uses its own shell; the human may also run it. Use it before merging the marketing branch into dev. Keep the project root clean on dev and finish pending merge/release/sync follow-ups first.

  1. Run cland publish <branch> --item T-n --dry-run. It selects branch-only commits (dev..<branch>), excluding inherited unreleased dev commits. Every path in every commit must match [publish] allow_paths in .cland/config.toml (exact paths or directory/** prefixes). The default permits site, brand docs, README, LICENSE and the Pages workflow. Mixed commits and merge commits refuse the whole operation; prepare a linear marketing branch instead. Never broaden the allow-list just to pass unrelated product work.
  2. Share the preview's source, commits, paths and resulting GitHub/dev tips through artifact.create in the item's topic. Ask the human through ask, with free: true, to approve this exact publication. Wait for approval, as for a release. If the source, dev, GitHub tips or configuration changes, preview and ask again.
  3. Run cland publish <branch> --item T-n. It cherry-picks onto GitHub's default branch, atomically fast-forwards that branch and site to one tip, merges both the source branch and public tip into dev, and syncs local refs and the territory. This sync preserves existing dev ancestry (D59); it does not replay open lanes. The hub records a durable publication landing, comments the public and dev tips on the item and leaves it open. Close the completed task separately; its publication landing then counts toward plan progress, with the stage’s QA and platform covers still required. Commit that tracker change on root dev and use cland push to publish it to the territory.
  4. If publication reports a failed follow-up or loses its push reply, preserve .cland/publish.tmp, repair the named obstruction and retry the same branch and item. The receipt pins the original tips; an unpublished retry must still satisfy the current allow-list. Once GitHub already has both tips, local-only recovery can finish after a policy change. Recovery never adds a second public history or duplicate tracker comment. Pending publication blocks release, ordinary sync, push, merge and offline tracker migration until recovered. A pre-push conflict leaves public and local refs untouched.

Publish creates no version or tag and makes no GitHub API calls. Ordinary cland sync retains its replay behavior; use publish's integrated sync for this marketing flow.

Coordinator publication procedure

Clause procedure (defaults coordinator):

  • Use release: cland release --item T-n --dry-run; ask notes/version approval, then cland release --item T-n. Human chooses version; Aides/stand-ins cite recorded choice. Marketing-only publication outside a release: cland publish <branch> --item T-n --dry-run; ask human approval, then cland publish <branch> --item T-n. Follow the release skill and recorded acceptance policy; never delegate release. cland sync imports from a clean root.

Clause procedure (project coordinator):

  • Use release: cland release --item T-n --dry-run, then ask for notes/version approval; once approved, cland release --item T-n. The human chooses the version; Aides/stand-ins cite a recorded choice. Follow recorded acceptance policy; never delegate release. cland sync imports from a clean root.

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.