agentleFS
akunzaiakunzai/agent-skills/skills/pr-workflow/SKILL.md

pr-workflow

Skill5 starsChanged 54 days agoMITView on GitHub

Use when preparing, creating, updating, or reviewing pull requests — enforcing pre-PR checks, atomic commits, dependency release links, and issue linkage (GitHub primary, non-GitHub referenced).

  • Commits and pushes

22 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use workflow and code-review.

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.
  • VercelSkill

    workflow

    Creates durable, resumable workflows using Vercel's Workflow SDK. Use when building workflows that need to survive restarts, pause for external events, retry on failure, or coordinate multi-step operations over time. Triggers on mentions of "workflow", "durable functions", "resumable", "workflow sdk", "queue", "event", "push", "subscribe", or step-based orchestration.

    Also in 4 projects that use this

  • GoogleSkill

    code-review

    Adversarial systems, concurrency, memory-model, and architecture code review protocol. Use this skill whenever asked to review code, PRs, diffs, or branches in Filament.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

11 rules. 0 are common, 6 are distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 19 edited versions we read keep these.

  • 6

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

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

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

    Base Sync: Before creating a feature branch or starting a new task, fetch and sync the default base branch (git fetch origin && git …

    line 15
  2. Distinctive:

    Clean Working Tree: Verify git status is clean before editing or opening PRs.

    line 16
  3. Distinctive:

    Verification: Run local tests and linters in the primary session rather than a subagent before opening or updating PRs.

    line 17
  4. Distinctive:

    If a named role is unavailable/unsupported or returns an explicit pre-execution dispatch/runtime error (for example capacity, rate limit, …

    line 25
  5. Distinctive:

    If a dispatch error does not reveal whether execution began, stop that dispatch and report the ambiguity.

    line 38
  6. Very specific:

    Stacked PRs: Prefer native stacks over hand-set feature-branch bases — GitHub gh stack (public preview.

    line 46
  7. Distinctive:

    Atomic Commits: Keep commits single-concern.

    line 55
  8. Very specific:

    Part-of: Use Part of #N or See #N for epic tracking and multi-PR tasks.

    line 61
  9. Very specific:

    No Tracked Issue: When a PR has no tracked issue, drop the Related Issue section entirely — never leave an unlinked Closes # or empty issue …

    line 63
  10. Distinctive:

    Use the worker-routing fallback above when the named summarizer is unavailable or unsupported.

    line 75
  11. Very specific:

    Before writing the request body, read the repo's docs/agents/pull-request.md (or equivalent: docs/agents/merge-request.md, native .github / …

    line 78

“Common” is measured: kept by at least 6 in 10 of the 19 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

PR Workflow

Standard operating procedure for preparing, opening, and managing Pull Requests (PR) and Merge Requests (MR). Default to GitHub (gh CLI); for other platforms, see platform-tools.md.

Preflight

  • Base Sync: Before creating a feature branch or starting a new task, fetch and sync the default base branch (git fetch origin && git checkout main && git pull --ff-only or branch directly off origin/main).
  • Clean Working Tree: Verify git status is clean before editing or opening PRs.
  • Verification: Run local tests and linters in the primary session rather than a subagent before opening or updating PRs; a failure needs their full output.

Worker routing

Request the cheapest capable model and lowest sufficient effort (low for routine); unsupported overrides inherit parent/configured defaults. Report requested/actual only from runtime metadata, else inherited/unknown.

If a named role is unavailable/unsupported or returns an explicit pre-execution dispatch/runtime error (for example capacity, rate limit, rejected model, or launch error), try one generic fallback that preserves the named worker's tools and permissions:

  • Log: exact approved local artifact only; reject unsafe input; report causes and events without fetching runs.
  • Evidence: caller's scenario only; stop what it started; no Git mutation or upload; report per-step expected, observed, verdict, and paths.

Otherwise use primary; a generic pre-execution failure also falls back to primary. Once a worker begins its assigned workload, its rejection or failure is final: no other worker, primary rerun, stronger model, or higher effort. If a dispatch error does not reveal whether execution began, stop that dispatch and report the ambiguity; do not retry or duplicate that workload.

Branching & Guardrails

  • Branch Guard: Stay on feature branches; never push or merge directly into the default base branch without explicit approval.
  • Base Discovery: Detect the base branch dynamically from repo evidence (origin/HEAD, PR target).
  • Fork Safety: Confirm with user before opening PRs on upstream or third-party forks.
  • Stacked PRs: Prefer native stacks over hand-set feature-branch bases — GitHub gh stack (public preview; official github/gh-stack skill), GitLab glab stack (experimental; bundled glab-stack skill). Hand-stacked: merge bottom-up, retargeting each PR onto the default branch first — closing keywords only fire there.

Commit Hygiene

  • Atomic Commits: Keep commits single-concern. Include release/changelog URLs on single dependency bumps.
  • Non-interactive: Pass GIT_EDITOR=true or --no-edit on git rebase/commit/merge to bypass GUI editor prompts.

Issue Linking

  • Auto-close: Reserve Closes #N, Fixes #N, or Resolves #N solely for PRs that auto-close the issue on merge.
  • Part-of: Use Part of #N or See #N for epic tracking and multi-PR tasks. GitHub matches those tokens even inside negation — never write Closes #N (or Fix/Resolve) next to an issue you must not close.
  • Commit messages: A closing keyword before #N in a commit message closes the issue once the commit reaches the default branch. GitHub and GitLab both accept a colon after the keyword, so a Conventional fix: #N ... subject closes #N. In commits for a multi-PR issue, keep the keyword off the number: fix(scope): describe change (#N).
  • No Tracked Issue: When a PR has no tracked issue, drop the Related Issue section entirely — never leave an unlinked Closes # or empty issue marker in the PR body.

PR Lifecycle (GitHub Primary)

For GitLab, Gitea, Azure DevOps, or Bitbucket, see platform-tools.md.

For a caller-selected remote failure log, classify and scope the source before delegation. The primary downloads only that log to a local artifact; the worker never fetches remote logs or discovers unrelated runs. Low-risk build, lint, and test artifacts may go to a named log-summarizer after caller review. Potentially sensitive logs require a sanitized artifact and residual-secret gate first; otherwise keep them in primary. Use the worker-routing fallback above when the named summarizer is unavailable or unsupported.

Before writing the request body, read the repo's docs/agents/pull-request.md (or equivalent: docs/agents/merge-request.md, native .github / GitLab templates) and follow its description shape and visual table, keying the visual to the user-visible change — a CLI or UI output change wins over an internal flowchart even when one exists. Where the table asks for a before/after pair, show both states, not just the after. Keep a multi-line text capture in a fenced block of its own under its Before or After label.

For a running-app capture of more than a step or two, hand a named evidence-collector the scenario (entrypoint, fixtures, steps, expected observations); on Cursor CLI keep it in primary. Judge the captures, mask flagged regions, then --attach.

Create PR

gh pr create --title "<type>(<scope>): <summary>" --body "<description>"

Update PR (Amend / Force-push)

gh pr edit <pr_number> --title "<updated_title>" --body "<updated_body>"

Guard body rewrites: fetch, edit, and write back a PR/MR/issue body in one set -euo pipefail block; never write an empty or shrunken body (test -s) — see body-rewrite.md.

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.