agentleFS
everrunseverruns/yolop/.agents/skills/ship/SKILL.md

ship

Skill18 starsChanged 13 days agoMITView on GitHub

Goal-oriented workflow for landing a requested change to yolop safely. Use when the user asks to ship, fix and ship, take a change through validation, or drive PR/CI/merge to completion.

194 other projects use a version of this, 1 of them edited.

Adopted by

Most recently taken up by

Used with

Projects with it also use ios-sync, open-gstack-browser 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.
  • Saqlain-811Skill

    ios-sync

    Regenerate the iOS debug bridge against the latest upstream gstack templates. (gstack)

    Also in 6 projects that use this

  • seek888Skill

    open-gstack-browser

    Launch GStack Browser — AI-controlled Chromium with the sidebar extension baked in. Opens a visible browser window where you can watch every action in real time. The sidebar shows a live activity feed and chat. Anti-bot stealth built in. Use when asked to "open gstack browser", "launch browser", "connect chrome", "open chrome", "real browser", "launch chrome", "side panel", or "control my browser". Voice triggers (speech-to-text aliases): "show me the browser".

    Also in 10 projects that use this

  • ayudaySkill

    setup-deploy

    Configure deployment settings for /land-and-deploy. Detects your deploy platform (Fly.io, Render, Vercel, Netlify, Heroku, GitHub Actions, custom), production URL, health check endpoints, and deploy status commands. Writes the configuration to CLAUDE.md so all future deploys are automatic. Use when: "setup deploy", "configure deployment", "set up land-and-deploy", "how do I deploy with gstack", "add deploy config".

    Also in 19 projects that use this

  • ShreyRaviSkill

    landing-report

    Read-only queue dashboard for workspace-aware ship. Shows which VERSION slots are currently claimed by open PRs, which sibling Conductor workspaces have WIP work likely to ship soon, and what slot /ship would pick next. No mutations — just a snapshot. Use when asked to "landing report", "what's in the queue", "show me open PRs", or "which version do I claim next". (gstack)

    Also in 9 projects that use this

  • zebbernSkill

    debug-fix

    Find and fix a bug or issue — from any source (GitHub issue, error message, user report, or observed behavior)

    Also in 4 projects that use this

  • ramybarsoumSkill

    document-release

    Post-ship documentation update. Reads all project docs, cross-references the diff, updates README/ARCHITECTURE/CONTRIBUTING/CLAUDE.md to match what shipped, polishes CHANGELOG voice, cleans up TODOS, and optionally bumps VERSION. Use when asked to "update the docs", "sync documentation", or "post-ship docs". Proactively suggest after a PR is merged or code is shipped.

    Also in 23 projects that use this

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 4 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 181 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

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

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

    Read knowledge/specs/shipping.md § Required Outcomes first — it owns the bar every shipped change must clear.

    line 14
  2. Very specific:

    Start from the goal and the changed risk surface, not from checklist order.

    line 21
  3. Very specific:

    The smoke test: Run the affected flow end to end.

    line 37
  4. Distinctive:

    Before pushing, reread the diff for duplication and accidental complexity you introduced, and fix it.

    line 42
  5. Distinctive:

    Stop and report only for blockers you cannot resolve alone: merge conflicts you cannot judge, missing credentials, ambiguous product …

    line 45
  6. Very specific:

    Write the body around functional change and impact — what changed, why, how it was validated, notable risks — using …

    line 78
  7. Distinctive:

    Every review comment gets an inline reply on its own thread and is marked resolved — including nits, low-confidence suggestions, and bot …

    line 95
  8. Distinctive:

    After the merge lands and main CI is green, close out with the shipped link: the merged PR URL and the merge commit on main.

    line 114

“Common” is measured: kept by at least 6 in 10 of the 181 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

Ship

Goal: land the requested change safely, with evidence, and merge only after CI is green.

Read knowledge/specs/shipping.md § Required Outcomes first — it owns the bar every shipped change must clear. This skill owns how to reach it.

"Fix and ship" means implement first, then switch into shipping mode.

Working the change

Start from the goal and the changed risk surface, not from checklist order. Review the delta (git diff origin/main...HEAD, git log origin/main..HEAD), confirm the requested behavior is actually implemented, then pick the smallest evidence that would convince a skeptical reviewer — targeted diff review, focused tests, then the checks in AGENTS.md for the surfaces you touched.

Two pieces of evidence are not interchangeable, and neither substitutes for the other:

  • The feature test. It must drive the changed behavior's real entry point — dispatch the command, call the handler, run the turn — not assert on a constructor or on adjacent code that still compiles. Changes to transcript output, live activity, or status values assert the terminal-independent presentation model; terminal-buffer tests are layout coverage, not proof of visible semantics.
  • The smoke test. Run the affected flow end to end. Agent-loop changes go through a live provider (doppler run -- cargo run -- --provider openai -p "<focused prompt>"); for offline-safe changes, --provider llmsim proving the binary still starts is enough.

Before pushing, reread the diff for duplication and accidental complexity you introduced, and fix it.

Stop and report only for blockers you cannot resolve alone: merge conflicts you cannot judge, missing credentials, ambiguous product intent, CI failures you cannot reproduce.

Security review

Mandatory for every change touching code, configuration, or infrastructure — perceived low risk does not excuse it. Yolop is a coding agent with disk and shell access on the user's host, so the categories are concentrated:

  • TM-FS — filesystem. The write blocklist still covers .git/, node_modules/, target/, dist/, build/, .next/, .venv/, venv/, .tox/, .gradle/ at any depth; reads stay unrestricted only inside the workspace root.
  • TM-BASH — shell execution. Timeouts and per-stream output caps survive. Any change to tools.rs must preserve the bounded execution model.
  • TM-LLM — prompt construction and key handling. Keys are read from process env only, and never logged or written to session JSONL.
  • TM-TOOL — capability registration. New capabilities respect the write blocklist.
  • TM-DEP — dependency risk. A new crate needs a one-line justification. everruns-* versions move together; mismatched versions are a soft API break.

For each relevant category, check the diff for injection (command, prompt, path traversal), data exposure, missing validation at trust boundaries, and resource exhaustion (unbounded loops, missing limits).

Record the result in the PR body under Security. Docs-only, comment-only, or test-only changes may state "No security-relevant code changes" with a one-line justification.

PR and merge

Write the body around functional change and impact — what changed, why, how it was validated, notable risks — using .github/pull_request_template.md.

Attach a Before / After with proof a reviewer can check — CLI output or a terminal recording (asciinema/VHS) of the agent run, and screenshots for any TUI change. Make screenshots 2x and lossless; the landing sharp-screenshots skill covers terminal captures and why remote captures come out blurry. State explicitly when a change has no observable behavior. Two sections are never omitted: Security (above) and Follow-ups (everything deferred, one line of rationale each, or "No follow-ups."). Default to implementing in-scope work rather than deferring it.

Record the knowledge decision explicitly: either the concepts you updated, or "No knowledge update required" with a reason.

Every review comment gets an inline reply on its own thread and is marked resolved — including nits, low-confidence suggestions, and bot comments. A reply is required even when the resolution is a pure code change.

Observe PR identity and CI status through separate structured fields. Capture gh pr view <PR> --json headRefOid --jq .headRefOid, then wait with gh pr checks <PR> --watch --fail-fast --interval 30 or gh run watch <RUN> --exit-status under the background task deadline. A terminal failure ends the watcher with a nonzero status. Recheck the PR head after the wait and immediately before merging; if it changed, validate the new head. gh pr checks output does not contain a head SHA, so never wait for a SHA to appear there. Preserve producer exit status and use fresh verification artifacts.

Merge with gh pr merge --squash after CI is green and a final comment sweep is clean. Emit the merge tool call in the same response as the green verdict, never announce the merge in a text-only response: a text-only response ends the turn and leaves the PR unmerged. Give async reviewer bots at least 2 minutes after CI turns green. Do not enable auto-merge — bots can post after the last push. After the merge lands, watch main CI for the merge commit and fix or revert promptly if it fails. After the merge lands and main CI is green, close out with the shipped link: the merged PR URL and the merge commit on main. Never end a ship run without handing the user the link to what shipped.

Approval

Asking to ship is the approval for shipping. Pushing the branch, opening the PR, and squash-merging it once the bar above is met are what this workflow exists to do, so run them without a request_approval pause and note the grant once with record_approval. Soft approval otherwise treats those as outward-facing and stops in front of them, which is how a run ends on "Squash-merging." and waits for a "do this" that adds nothing: the gate that actually protects the merge is green CI and a clean sweep, not a second yes.

The grant covers this change through its merge and nothing else. Force-pushing or rewriting Git history requires explicit human approval for that specific action, including when approval mode is off. A recorded approval cannot create that grant. Reverting on main or touching another PR also needs separate approval.

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.