agentleFS
SDSLeonSDSLeon/lightcode/.agents/skills/address-pr-feedback/SKILL.md

address-pr-feedback

Skill121 starsChanged 5 days agoApache-2.0View on GitHub

Collects, validates, triages, and resolves review feedback on a Poracode pull request, then replies in each thread. Use when asked to "address review comments", "handle PR feedback", "respond to the review", "fix the review threads", or "triage the PR comments". Not for self-review before opening a PR (use deep-review) or for reviewing someone else's PR.

  • Commits and pushes

6 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

12 rules. 0 are common, 7 are distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 7

    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
  • could be automatic: a setting could enforce it
  1. Distinctive:

    If the user asked only for triage, stop after the triage table: do not edit, push, or post replies.

    line 10
  2. Very specific:

    Check out the PR branch (gh pr checkout <n>) and gather all three feedback channels.

    line 14
  3. Very specific:

    If hasNextPage is true, page with an after: <endCursor> argument.

    line 35
  4. Distinctive:

    Read the current code at the referenced path, not only the diff hunk.

    line 49
  5. Very specific:

    Check the claim against repo rules (AGENTS.md, .agents/docs/).

    line 51
  6. Distinctive:

    Ask the user about "Needs owner decision" items before acting on them.

    line 67
  7. Distinctive:

    Keep fixes scoped to the feedback.

    line 72
  8. Very specific:

    Run the verify skill.

    line 73
  9. Very specific:

    Commit with the commit skill conventions and push to the PR branch with a plain git push.

    line 74
  10. Distinctive:

    Do not resolve reviewers' threads or dismiss reviews.

    line 89
  11. Distinctive:

    Every unresolved item has a verdict in the triage table.

    line 95
  12. Distinctive:

    Every thread has a reply with a SHA or a reason, and reviewers are re-requested.

    line 97

“Common” is measured: kept by at least 6 in 10 of the 5 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

Address PR feedback

The author's response to other people's review. Every item gets a verdict backed by evidence, accepted items get fixed and verified, and every thread gets a reply.

If the user asked only for triage, stop after the triage table: do not edit, push, or post replies.

1. Collect everything

Check out the PR branch (gh pr checkout <n>) and gather all three feedback channels; reviewers use all of them.

Inline review threads, with resolution state:

gh api graphql -F owner='{owner}' -F repo='{repo}' -F pr=<n> -f query='
query($owner: String!, $repo: String!, $pr: Int!) {
  repository(owner: $owner, name: $repo) {
    pullRequest(number: $pr) {
      reviewThreads(first: 100) {
        pageInfo { hasNextPage endCursor }
        nodes {
          id isResolved isOutdated path line originalLine
          comments(first: 50) { nodes { databaseId author { login } body url } }
        }
      }
    }
  }
}'

If hasNextPage is true, page with an after: <endCursor> argument; outdated threads have no line, so use originalLine.

Review summaries and conversation comments:

gh pr view <n> --json reviews,comments,headRefName,url

Skip resolved threads unless a later comment reopens the point. Treat outdated threads as live until you confirm the current code addresses them.

2. Validate each item against the code

A reviewer comment is a claim, not a fact. Before accepting:

  • Read the current code at the referenced path, not only the diff hunk; the line may have moved.
  • For behavioral claims, reproduce: write or run a failing test, or drive the app via interactive-testing.
  • Check the claim against repo rules (AGENTS.md, .agents/docs/). A suggestion that breaks provider isolation, i18n, or versioning rules is not accepted as written.

3. Triage

Produce one table before changing anything:

#Source (thread/review/comment link)ClaimVerdictEvidence / plan

Verdicts:

  • Accept: valid, fix as suggested.
  • Accept with different fix: the problem is real but the suggested change is wrong or narrower than needed; describe the alternative.
  • Reject: the claim does not hold; cite the code, test, or doc that shows it.
  • Needs owner decision: product, scope, or design trade-off the author cannot settle alone; state the options.

Ask the user about "Needs owner decision" items before acting on them.

4. Fix

  • Fix accepted items. Add or update a test wherever behavior changed; the test is the proof for the reply.
  • Keep fixes scoped to the feedback. Unrelated improvements go in a separate PR.
  • Run the verify skill.
  • Commit with the commit skill conventions and push to the PR branch with a plain git push. Do not force-push unless the user asks; it breaks reviewers' "changes since last review" view.

5. Reply and re-request

Reply in each thread, using the databaseId of the thread's first comment:

gh api repos/{owner}/{repo}/pulls/<n>/comments/<comment-id>/replies -f body='Fixed in <sha>: <what changed>. Test: <test name or n/a>'
  • Fixed: Fixed in <sha>: <what>. Test: <test or n/a>
  • Not changed: Not changed: <reason and evidence>

Answer review summaries and conversation comments with one gh pr comment <n> --body-file <file> that covers each point.

Do not resolve reviewers' threads or dismiss reviews; the reviewer decides when a point is settled.

Re-request review from each reviewer who left feedback: gh pr edit <n> --add-reviewer <login>.

Done when

  • Every unresolved item has a verdict in the triage table.
  • Accepted items are fixed, tested where behavior changed, verified, and pushed.
  • Every thread has a reply with a SHA or a reason, and reviewers are re-requested.
  • The report includes the triage table, pushed SHAs, checks run, and any items awaiting an owner decision.

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.