
autoship
Runs a changesets npm release through the version PR, CI publish, and registry verification. Use when asked to "release this package", "autoship", "merge Version Packages", or diagnose a release that did not publish.
- Installs
- 627
- Last 24h
- —
- Trending rank
- —
- Security
- Not audited
Install
Works with every agent the skills CLI supports.
npx skills add https://github.com/mblode/agent-skills --skill autoship- 01
Install
Run the command in your project. The CLI asks which agents to install for.
- 02
Describe the task
Ask your agent for something this skill covers. It loads autoship automatically when relevant.
- 03
Review the result
The agent follows the SKILL.md steps, scripts and references bundled with the skill.
SKILL.md
The instructions your agent reads when the skill is triggered.
Autoship
Drive an npm release end to end: changeset, fix loop, push, CI watch, Version Packages PR merge, publish watch, npm verification.
- IS: the full release pipeline for an existing changesets-based npm package, from writing the changeset file to confirming the new version on the registry, plus diagnosing why a release run did not version or publish.
- IS NOT: opening a feature PR (use
pr-creator), monitoring a feature PR for reviews, conflicts, or CI (usepr-babysitter), general build or type fixes outside a release flow, or scaffolding a new package (usescaffold-cli, which hands off to autoship for the first release).
The Release Loop
One workflow, two successive runs. Misreading it as two workflows causes most autoship mistakes.
- Push a commit containing a pending
.changeset/*.mdfile to the default branch. - The release workflow runs:
changesets/actionsees pending changesets, runschangeset versionin CI, and opens or updates a PR on branchchangeset-release/<default-branch>(title "Version Packages", suffixed "(next)" in pre mode) carrying thepackage.jsonbump andCHANGELOG.mdentry. - Merge that PR once every check is green.
- The same workflow runs again. With no pending changesets left, the action runs its publish script (
changeset publish), which publishes to npm and, by default, pushes the git tag and creates a GitHub release.
The local job ends at "push the changeset file". CI owns versioning and publishing.
The action has two live majors with different input names: @v1 takes publish:, @v2 takes publish-script:. Read the uses: line before diagnosing a run that versioned but never published.
Reference Files
| File | Read when |
|---|---|
references/changeset-and-commit.md | Steps 1-3: writing the changeset file, discovering and running gates non-interactively, staging the release commit |
references/ci-polling.md | Step 4 and Step 5 watches: the Monitor tool, the commit watch script, failure classification, the Changeset Status check |
references/version-pr-and-publish.md | Once CI is green: workflow shape, finding and merging the Version Packages PR, the publish run, npm verification, publish failure diagnosis |
evals/evals.json | Only when changing this skill; never loads during a release |
Intent Map
| Intent | Steps | Notes |
|---|---|---|
| Full autoship (ship / release / publish) | 1-5 | Default entry point. End to end through publish, no intermediate prompts |
| Create changeset only | 1 | Stage a release without pushing |
| Fix gates and push | 1-3 | Changeset, fixes, commit, no CI watch |
| Watch CI only | 4-5 | Changeset already pushed |
| Merge Version Packages PR only | 4b-5 | CI already green; merges once preconditions hold |
| Fix gates only | 2 | Inside a release flow; no changeset needed |
| Diagnose a release that did not publish | read-only | references/version-pr-and-publish.md (PR-wait timeout checks, publish failure table) |
"Ship it" with no npm release context routes to pr-creator.
Safety Tiers
Invoking autoship is standing consent for the full release flow. Do not pause mid-flow to re-confirm; gate risky steps on objective preconditions instead.
- Green (execute directly): reads: CI and PR state,
npm view, pending changesets,package.jsonscripts, git history and status. - Yellow (announce in one line, then execute): writing changeset files, running lint/type/test/format fixers,
git add/commit/push, starting watches, and merging the Version Packages PR once its identity is confirmed and every check passes. - Red (explicit confirmation required): force-pushes, history rewrites, repository settings changes, any destructive git operation.
Workflow
Copy this checklist to track progress:
Autoship progress:
- [ ] Step 1: Create changeset (default patch)
- [ ] Step 2: Fix lint, types, tests, format
- [ ] Step 3: Commit and push the changeset
- [ ] Step 4a: Watch CI on the pushed commit
- [ ] Step 4b: Find and merge the Version Packages PR
- [ ] Step 5: Watch the publish run, verify on npm
Step 1: Create changeset (default patch)
- Load
references/changeset-and-commit.md. - Inspect pending changesets and their package coverage. Reuse those covering the requested release; add one only for uncovered changes. Ask only if unrelated pending releases make the publish scope ambiguous.
- Default to
patch;minorormajoronly on explicit user instruction. - Write the file directly (the interactive prompt needs a TTY); the summary is user-facing changelog text inferred from
git log --oneline -10. pnpm changeset statusvalidates the file: a misspelled package name fails here instead of in CI.
Step 2: Fix lint, types, tests, format
- Discover commands from
package.jsonscripts (check,lint,typecheck,test,format,fix); in non-Node repos checkMakefile,Cargo.toml,pyproject.toml,go.mod. - Run lint, typecheck, test, format. After any code change, re-run from the first gate: a type fix routinely breaks lint, and a lint autofix can break a test.
- Scope auto-fixers to changed files where supported, then check
git status: broadfix/formatscripts reformat files outside the change (MDX is a frequent casualty). Undo only fixer changes introduced by this run, preserving pre-existing edits in the same files. - Cap the loop at 5 fix iterations per gate, reporting the remaining error count each pass; then stop and report the gate, remaining error count, and last error output.
Step 3: Commit and push the changeset
- Stage the changeset file and in-scope fixes by explicit path; sweep
git status --porcelainfor hook artifacts (a rootschema.gqlis a known one) before committing. - Commit (
chore: add <type> changeset for <package>) and push. - The pushed commit must still contain
.changeset/*.md(first Gotcha).
Step 4a: Watch CI on the pushed commit
- Load
references/ci-polling.md. - Start the commit watch on the pushed SHA. It emits a line per state change and one
TERMINAL:line when every run for that commit (CI and the release workflow) completes. An idle first poll is normal; runs take time to queue. - On failure, classify from
gh run view <id> --log-failed: flaky or infra getsgh run rerun <id> --failed(max 3); a real failure gets a fix, commit, push, and a fresh watch on the new SHA.
Step 4b: Find and merge the Version Packages PR
- Load
references/version-pr-and-publish.md. - Find the open PR whose head is
changeset-release/<default-branch>. If absent, run the PR-wait watch (10-minute cap). - Merge only when all three hold:
- Head branch is
changeset-release/<default-branch>(title normally "Version Packages", butpr-titlecan rename it). Never merge any other PR. gh pr checks <n> --json name,bucketreportsbucket: passfor every check.gh pr view <n> --json mergeablereportsMERGEABLE(onUNKNOWN, wait and re-query).
- Head branch is
- Announce ("Merging Version Packages PR #N: <package>@<version>"), then
gh pr merge <n> --squash --delete-branch, or the merge method the repo allows. Any failed precondition: stop and report, never merge.
Step 5: Watch the publish run, verify on npm
- Merging triggers the same workflow again; with no pending changesets it publishes.
- Take the merge SHA (
gh pr view <n> --json mergeCommit --jq .mergeCommit.oid) and run the commit watch on it. - On failure: read the log, match it against the publish failure table in
references/version-pr-and-publish.md, report, and stop. Publish failures are never retried blind; every cause on that table needs a config or settings change. - On success:
npm view <package>@<version> versionmust print the merged version, andnpm view <package> dist-tagsmust showlatestpointing at it (unless pre mode). Stop remaining watches and report both outputs; they are the completion evidence.
Failure Recovery
Stop and report on a gate still failing after 5 iterations or any failed merge precondition; never override failing checks or resolve conflicts in the bot PR. Diagnosis lives with each stage: CI and Changeset Status failures in references/ci-polling.md, a missing Version Packages PR and publish failures in references/version-pr-and-publish.md.
Gotchas
- Never run
pnpm changeset versionlocally. It consumes.changeset/*.md, so the pushed commit has no pending changeset, "Changeset Status" fails, and no Version Packages PR opens. Recovery is reverting the bump, not rerunning CI. - Never run
pnpm publishornpm publishdirectly. It bypasses changesets, skips the changelog and tag, and leaves the Version Packages PR describing an already-shipped version, which then fails with "You cannot publish over the previously published versions". - Never hand-edit
CHANGELOG.mdor thepackage.jsonversion. CI generates both in the Version Packages PR; local edits make the bot PRCONFLICTING. changesets/action@v2renamed every input (publishtopublish-script,versiontoversion-script,titletopr-title,committocommit-message,branchtopr-base-branch). The old names are ignored with only a warning, so a@v2workflow still usingpublish:opens the Version PR and then completes green without publishing anything.setup-nodewithnode-version: 22ships npm 10.9.x. Trusted publishing needs npm 11.5.1 or later, sochangeset publishfailsENEEDAUTHeven withid-token: writeset (pnpm hands the upload to the npm CLI onPATH). Use Node 24 or addnpm install -g npm@latestbefore publishing.- A workflow authenticating with an
NPM_TOKENsecret that holds a classic token failsENEEDAUTH: npm revoked every classic token on 9 December 2025. Move to trusted publishing rather than minting a new token; write-capable granular tokens expire within 90 days. - npm does not validate a trusted publisher when you save it. A workflow filename that differs from
.github/workflows/<file>.ymlby case or extension surfaces only at publish time, asENEEDAUTH. - New personal repos block Actions from opening PRs. The release run fails with "GitHub Actions is not permitted to create or approve pull requests" and no Version PR appears. Fix: Settings, Actions, General, "Allow GitHub Actions to create and approve pull requests" (an org-level setting can override it). Red tier: report it, do not change settings unasked.
gh pr list --json headBranchandgh pr checks --json conclusionare invalid fields and error. UseheadRefNameandbucket.- The Monitor tool's default
timeout_msis 300000 (5 minutes). A CI run that outlasts it kills the watch with a timeout that looks like silence. Passtimeout_ms: 3600000for CI watches. - Poll every 30 seconds or slower. Faster loops burn the GitHub API rate limit (
gh api rate_limit --jq .resources.core.remaining) and stall the flow mid-release. git add -Acommits pre-commit-hook artifacts and fixer churn into the release commit. Stage explicit paths.majorwithout explicit instruction signals breaking changes to every consumer. Default topatch.
Related Skills
scaffold-cli: scaffolds a new TypeScript package with changesets and OIDC publishing, then hands off to autoship for its first release.pr-creator: opens feature PRs. Autoship merges only the bot-opened Version Packages PR.pr-babysitter: watches feature PRs (reviews, conflicts, CI). Autoship watches release CI only; never babysit a Version Packages PR autoship is driving.
FAQ
What does the autoship skill do?
Runs a changesets npm release through the version PR, CI publish, and registry verification. Use when asked to "release this package", "autoship", "merge Version Packages", or diagnose a release that did not publish.
How do I install autoship?
Run npx skills add https://github.com/mblode/agent-skills --skill autoship in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/autoship/ (personal) or .claude/skills/autoship/ (project).
Which AI agents can use autoship?
Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.
Is autoship safe to install?
No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.
Who maintains autoship?
It is published by mblode in mblode/agent-skills. It has 627 recorded installs.
More from mblode/agent-skills
ui-animation
mblode/agent-skills
Builds, reviews, and measures UI motion, including springs, gestures, scroll effects, curve fitting from recordings, and sparse interface sound. Use when asked to "add animation", "match this easing", "reverse engineer this motion", "add a click sound", or find animation opportunities.
multi-tenant-architecture
mblode/agent-skills
Designs tenant isolation, hostname routing, custom-domain lifecycle, and plan limits on Cloudflare or Vercel. Use when asked to "isolate tenant data", "support custom domains", "build a white-label platform", or assess PSL registration. For general module structure use codebase-architecture; for SEO content use seo.
ui-design
mblode/agent-skills
Designs and builds React/Next/Tailwind UI, audits visual and interaction defects, and stress-tests components with worst-case data. Use when asked to "build a landing page", "extract our design system", "add dark mode", "make this responsive", "remove UI slop", "audit this component", "stress-test this component", "break this UI", or "show me the worst case".
typography-audit
mblode/agent-skills
Audits font loading, type scales, measure, spacing, OpenType, and rendered punctuation with 78 scoped rules. Use when asked to "audit typography", "fix the fonts", or "review my type system". For a new visual direction use ui-design Direction; for general UI defects use ui-design Audit.
agent-skills-creator
mblode/agent-skills
Creates and improves portable Agent Skills with a validator, routing scenarios, and evidence-based keep, cut, merge, or retire decisions. Use when asked to "write a skill", "update all skills", "audit my SKILL.md", "remove redundant instructions", or fix skill triggering.
docs-writing
mblode/agent-skills
Writes and audits technical docs with Diataxis type gating, runnable examples, and 51 documentation rules. Use when asked to "review my docs", "write an API reference", or improve a documentation page. For a whole README use readme-creator; for agent instructions use agents-md.
Popular ai & agents skills
google-agents-cli-publish
google/agents-cli
This skill should be used when the user wants to "publish an agent", "publish my ADK agent", "register an agent with Gemini", "publish to Gemini" (formerly Gemini Enterprise), or needs guidance on the agents-cli publish gemini command. Also use when the user wants to "manage agents in Agent Registry", "list/update/delete registered agents", or "register an MCP server". Covers ADK vs A2A registration modes, programmatic and interactive usage, flag reference, auto-detection from deployment metadata, Agent Registry fleet management, and troubleshooting. Part of the agents-cli skills suite. Do NOT use for deployment (use google-agents-cli-deploy).
google-agents-cli-eval
google/agents-cli
This skill should be used when the user wants to "run an evaluation", "evaluate my agent", "evaluate my ADK agent", "write an eval dataset", "analyze eval failures", "compare eval results", "optimize agent", or needs guidance on the Gemini platform eval methodology and the Quality Flywheel. Covers eval metrics, dataset schema, LLM-as-judge scoring, and common failure causes. Applies to any agents-cli project, whatever framework the agent is written in. Do NOT use for agent API code patterns (ADK: use google-agents-cli-adk-code), deployment (use google-agents-cli-deploy), or project scaffolding (use google-agents-cli-scaffold).
skill-creator
anthropics/skills
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
codex-pet
prime-skills/runcomfy-agent-skills
Codex Pet generator on RunComfy. Build a Codex-compatible Codex Pet spritesheet.webp + pet.json from a single reference image, drop it into ${CODEXHOME:-$HOME/.codex}/pets/ / and Codex picks it up as a custom Codex Pet next to the 8 built-ins. This skill produces the exact Codex Pet atlas Codex expects (1536x1872 PNG/WebP, 8 cols x 9 rows, 192x208 cells, 9 animation states — idle, running-right, running-left, waving, jumping, failed, waiting, running, review). Calls OpenAI GPT Image 2 edit ONCE via the local RunComfy CLI as runcomfy run openai/gpt-image-2/edit to produce a canonical Codex Pet pose, then assembles all 9 animation rows programmatically with ImageMagick micro-transforms — no Codex Pro, no $imagegen, no OPENAIAPIKEY required, only RUNCOMFYTOKEN. Triggers on "codex pet", "create codex pet", "make codex pet", "hatch codex pet", "/hatch image", "desktop pet codex", "codex pets", "spritesheet.webp", or any explicit ask to build a custom pet for OpenAI Codex.
full-output-enforcement
leonxlnx/taste-skill
Overrides default LLM truncation behavior. Enforces complete code generation, bans placeholder patterns, and handles token-limit splits cleanly. Apply to any task requiring exhaustive, unabridged output.
executing-plans
obra/superpowers
Use when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available