agentleFS
holon-runholon-run/holon/skills/github-pr-fix/SKILL.md

github-pr-fix

Skill194 starsChanged 4 days agoApache-2.0View on GitHub

Fix a GitHub pull request by addressing feedback or CI failures, pushing changes, and publishing replies.

  • Reads credentials

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

14 rules. About 0 are common, 10 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈10

    Distinctive · estimate

    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. Distinctive:

    Use this skill when you need to remediate an existing pull request: diagnose what is broken, apply targeted fixes, push commits, and reply …

    line 10
  2. Distinctive:

    Do not create fixed output files unless the caller or an integration explicitly requires them.

    line 33
  3. Very specific:

    If no manifest is provided, collect PR context directly with gh:

    line 44
  4. Very specific:

    Do not assume fixed github/*.json files.

    line 67
  5. Very specific:

    If ${GITHUB_CONTEXT_DIR}/manifest.json exists, use it.

    line 74
  6. Distinctive:

    Use existing review threads/comments to avoid duplicate or stale responses.

    line 86
  7. Distinctive:

    Follow the developer agent's recorded worktree, branch, and PR preferences.

    line 91
  8. Distinctive:

    Run relevant verification commands.

    line 93
  9. Distinctive:

    Commit and push before posting review replies.

    line 94
  10. Very specific:

    Use direct gh api reply operations:

    line 98
  11. Distinctive:

    Do not mark issues fixed without verification evidence.

    line 115
  12. Distinctive:

    If verification is partial, state exact limits and risk.

    line 116
  13. Distinctive:

    Keep review replies concrete: what changed, where, and any remaining risk.

    line 118
  14. Distinctive:

    If replies are planned but not published, the run is not successful.

    line 128

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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

GitHub PR Fix Skill

Summary

Use this skill when you need to remediate an existing pull request: diagnose what is broken, apply targeted fixes, push commits, and reply on review threads.

When To Use

  • Fixing CI failures on an existing PR
  • Addressing requested changes or review feedback
  • Updating the PR branch and publishing review replies

Do Not Use

  • Opening a brand-new PR from an issue
  • Performing a review-only pass without code changes
  • Project-wide triage or roadmap planning

Prerequisites

  • gh CLI authentication is required.
  • GITHUB_TOKEN/GH_TOKEN must allow PR read-write operations.

Runtime Paths

  • GITHUB_OUTPUT_DIR: optional caller-provided output artifacts directory.
  • GITHUB_CONTEXT_DIR: context directory (default ${GITHUB_OUTPUT_DIR}/github-context).
  • Do not create fixed output files unless the caller or an integration explicitly requires them.

Inputs (Manifest-First)

Preferred input when already available:

  • ${GITHUB_CONTEXT_DIR}/manifest.json

Optional inputs:

  • Any artifact listed as status=present in manifest.artifacts[].

If no manifest is provided, collect PR context directly with gh:

gh pr view <pr_number> --repo <owner/repo> --json number,title,body,state,url,baseRefName,headRefName,headRefOid,author,createdAt,updatedAt,mergeable,reviews,changedFiles,additions,deletions
gh pr view <pr_number> --repo <owner/repo> --json files
gh api repos/<owner>/<repo>/issues/<pr_number>/comments --paginate
gh api graphql -f query='
  query($owner:String!, $repo:String!, $number:Int!) {
    repository(owner:$owner, name:$repo) {
      pullRequest(number:$number) {
        reviewThreads(first:100) {
          nodes {
            isResolved
            comments(first:100) {
              nodes { id body path line author { login } }
            }
          }
        }
      }
    }
  }' -F owner=<owner> -F repo=<repo> -F number=<pr_number>

Do not assume fixed github/*.json files. Resolve available context through artifact metadata (id, path, status, description).

Workflow

1. Collect context

  • If ${GITHUB_CONTEXT_DIR}/manifest.json exists, use it.
  • Otherwise, collect PR metadata, files, comments, and review threads directly with gh.

2. Diagnose and prioritize

Prioritize in this order:

  1. Build/compile failures
  2. Failing tests and runtime regressions
  3. Type/import/module errors
  4. Lint/style issues
  5. Non-blocking refactor suggestions

Use existing review threads/comments to avoid duplicate or stale responses.

3. Implement fixes

  • Apply minimal targeted fixes for blocking issues first.
  • Follow the developer agent's recorded worktree, branch, and PR preferences; ask the operator when a relevant preference is not recorded.
  • Run relevant verification commands.
  • Commit and push before posting review replies.

4. Publish review replies

Use direct gh api reply operations:

gh api repos/<owner>/<repo>/pulls/<pr_number>/comments \
  -X POST \
  -F in_reply_to=<comment_id> \
  -f body='Thanks, fixed in the latest commit.'

5. Finalize delivery

Report the result in the normal agent delivery. If a caller explicitly requires machine-readable output, write only the requested artifacts under ${GITHUB_OUTPUT_DIR} and include their paths in the delivery.

Remediation Standards

  • Do not mark issues fixed without verification evidence.
  • If verification is partial, state exact limits and risk.
  • Defer non-blocking large refactors with explicit rationale.
  • Keep review replies concrete: what changed, where, and any remaining risk.

Completion Criteria

A successful run requires all of the following:

  1. Blocking fixes are committed and pushed to the PR branch.
  2. Replies planned for this run are published successfully.
  3. The agent delivery records what changed, which replies were posted, and any remaining risks.

If replies are planned but not published, the run is not successful.

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.