agentleFS
frostneyfrostney/knips/.agents/skills/create-pr/SKILL.md

create-pr

Skill1 starsChanged 54 days agoMITView on GitHub

Publishes a completed change as a templated draft pull request, reconciles PR-only metadata and readiness state, waits for CI, and marks it ready. Use when the user runs /create-pr.

  • Commits and pushes

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 178 edited versions we read keep these.

  • 11

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When the branch belongs to a native GitHub stack, read ../git-workflow/references/github-stacks.md.

    line 21
  2. Distinctive:

    Stop if there are no relevant changes or commits ahead of the remote base.

    line 29
  3. Distinctive:

    If currently on the base branch, create a focused branch named from the issue or change.

    line 31
  4. Distinctive:

    Verify that code review, observed behavior, and the repository's project gate all passed on the exact current change.

    line 39
  5. Distinctive:

    Stop when required review or behavior evidence is missing, stale, failed, or unverified without explicit acceptance.

    line 44
  6. Distinctive:

    Commit uncommitted work with a concise Conventional Commit subject.

    line 54
  7. Distinctive:

    Only after the publication checks pass, push an ordinary branch normally and set its upstream when needed, then open one draft PR against …

    line 71
  8. Distinctive:

    Run the PR-specific phase.

    line 77
  9. Distinctive:

    If the PR-specific phase or CI exposes a repository, code, behavior, test, documentation, generated-artifact, or compatibility gap, keep …

    line 81
  10. Distinctive:

    If the host cannot passively await a subprocess, report the unsupported capability instead of polling through model turns.

    line 92
  11. Distinctive:

    Keep the PR in draft while any requirement for marking it ready or CI check is pending or failing.

    line 94

“Common” is measured: kept by at least 6 in 10 of the 178 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

Create PR

The request authorizes the repository's declared gates, relevant commits, PR metadata updates, one ordinary draft pull request or the confirmed native stack layers owned by the change, and their transitions to ready for review. It does not authorize implementation fixes or unrelated changes.

When the branch belongs to a native GitHub stack, read ../git-workflow/references/github-stacks.md. The request then authorizes submission and metadata reconciliation for the confirmed stack layers owned by the current change, not unrelated branches.

  1. Inspect the working tree, staged diff, recent commits, remote default branch, stack topology when applicable, and any existing remote head. Preserve unrelated local work.
  2. Stop if there are no relevant changes or commits ahead of the remote base. Continue without an empty commit when the work is already committed.
  3. If currently on the base branch, create a focused branch named from the issue or change.
  4. Establish the PR claim and publication requirements before publishing anything. Read the explicit request, linked issue or confirmed mini-spec, applicable project instructions, product docs, ADRs or durable decisions, the nearest DEFINITION_OF_READY.md, and the completion evidence supplied by the implementation workflow. If no source states the claim, reconstruct the narrowest claim supported by the commits and diff and label it as inferred.
  5. Verify that code review, observed behavior, and the repository's project gate all passed on the exact current change. An explicitly accepted limitation must name the unverified behavior. Reuse a current project-gate result rather than running it again. If the gate result is missing or stale, run the declared gate once without weakening it.
  6. Stop when required review or behavior evidence is missing, stale, failed, or unverified without explicit acceptance. Stop when the project gate fails. Report the exact missing evidence or failure and return the change to its implementation workflow. Do not test delivered behavior, inspect source as a substitute, or fix implementation, test, documentation, generated-artifact, compatibility, or CI gaps in this workflow. The only publication exception is an explicit user request for a draft PR to obtain a preview environment that does not otherwise exist. That PR must remain draft until behavior is tested on a preview tied to its exact revision.
  7. Stage only relevant files, excluding secrets and unrelated local work. Commit uncommitted work with a concise Conventional Commit subject. Never amend and never skip hooks. Preserve already-published history and add a new commit for any correction.
  8. Title each pull request with a Conventional Commit subject covering the whole change, since the squash merge makes that title the commit subject on the base branch. Fill the matching PR template for each submitted layer and preserve its structure. If none exists, use Summary, Testing, and Linked issues. Keep the body proportional to the change. Before writing it, search open and closed issues and recent sibling sessions or adjacent branches when available for related findings and duplicates. Put each closing keyword on its own line as Closes #N, and only on the layer that completes that issue. Upload local screenshots and videos with the GitHub CLI's --attach flag (gh pr create --attach '<file>#<alt text>', or gh pr edit and gh pr comment with the same flag) so the body links the uploaded asset; never commit evidence media to the repository. Put non-media evidence such as probe output or a plan inside a collapsed <details> block in the body or a comment.
  9. Only after the publication checks pass, push an ordinary branch normally and set its upstream when needed, then open one draft PR against the remote default. For a verified native stack, use gh stack submit only after every submitted layer has current completion evidence; only guarded official stack operations may rebase or push with force-with-lease. Preserve bottom-to-top topology and keep each layer draft.
  10. Run the PR-specific phase. Compare the actual PR diff, body, links, metadata, committed tests and documentation, supplied completion evidence, and facts that exist only after publication. Correct metadata-only gaps without a commit.
  11. If the PR-specific phase or CI exposes a repository, code, behavior, test, documentation, generated-artifact, or compatibility gap, keep the PR draft and return the exact failure to the implementation workflow. Do not fix it here. A later caller may update the PR after the appropriate implementation, review, behavior-testing, and project-gate loop passes.
  12. Invoke delivery-wait's foreground wait checks-terminal operation with the repository, pull request number, exact head, one --check per required check context, absolute deadline, an optional --state path, and --json; the harness passively awaits it without model heartbeats. When it returns, inspect unsuccessful logs. Report a validated code or behavior failure under step 11 rather than fixing it. If the host cannot passively await a subprocess, report the unsupported capability instead of polling through model turns.
  13. Keep the PR in draft while any requirement for marking it ready or CI check is pending or failing. A PR opened to obtain a preview returns to the implementation workflow for black-box testing and cannot become ready in this run. If an expected pull-request workflow produces no run, read the PR's mergeable_state first: a dirty PR gets no pull-request runs, and no retrigger can help until the conflict is resolved. Continue monitoring pending checks; if they cannot reach a terminal result during the run, report their current state. Stop and report the exact blocker for a material decision, unrelated failure, unavailable external service, or validation that cannot be performed safely.
  14. Once the PR is missing nothing required by the publication and PR-specific phases and all applicable CI is observed green for its exact head, mark it ready for review. Return every affected URL, native stack order when applicable, final states, metadata changes, supplied completion evidence, and observed readiness and CI evidence.

Download now (ZIP)

  • Other versions196 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 2

    1. frostney/pascal-mcp-sdk★ 232% same
    2. frostney/known-good-route★ 027% same

    Wrote their own version · 194

    1. microsoft/vscode★ 194ksame name
    2. vercel/next.js★ 143ksame name
    3. penpot/penpot★ 62ksame name
    4. remix-run/react-router★ 57ksame name
    5. moeru-ai/airi★ 51ksame name
    6. ToolJet/ToolJet★ 41ksame name
    7. ueberdosis/tiptap★ 39ksame name
    8. BabylonJS/Babylon.js★ 26ksame name
    9. kortix-ai/suna★ 20ksame name
    10. ryoppippi/ccusage★ 19ksame name

    184 more not listed

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • Used alongsideProjects with it also use act-on-feedback, create-release-note and 6 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

      act-on-feedback

      Act on user feedback attached to the current session. Use when the user submits feedback on the session's changes via the Submit Feedback button.

      Also in 3 projects that use this

    • ModelEngine-GroupSkill

      create-release-note

      从 PR/commit 中自动生成结构化的 Release Notes,并可选创建 GitHub Draft Release。支持 x.y.0 版本合并前一 minor 系列的 release notes。当用户要求生成发布说明、创建 release notes 时触发。参数为版本号和可选的上一版本号。

      Also in 3 projects that use this

    • MicrosoftSkill

      generate-run-commands

      Generate or modify run commands for the current session. Use when the user wants to set up or update run commands that appear in the session's Run button.

      Also in 3 projects that use this

    • Hey-DigaSkill

      review-pr-comments

      Review and apply Claude's PR feedback interactively. Use after receiving a Claude Code review on a PR to triage and implement feedback.

      Also in 4 projects that use this

    • RisorseArtificialiSkill

      issue-reply

      Reply to GitHub issue discussions as the maintainer — fetch the issue and its thread, brief the user on the discussion and the open questions, verify technical claims against the code before endorsing them, draft the reply, and publish only after explicit approval. Use when the user says "rispondi alle issue", "respond to the issues", "guardiamo la #N", "reply to the comments", or hands over a list of issues to answer. The commenting counterpart of issue-triage, which stays read-only.

      Also in 3 projects that use this

    • aakash-dharSkill

      debug-code

      Debugs code issues by analyzing errors, tracing execution flow, identifying root causes, and suggesting fixes. Use when the user says "this isn't working", "I'm getting an error", "why is this broken?", or pastes an error message.

      Also in 3 projects that use this

  • In this repository19 other files it gives its agents.

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.