agentleFS
MicrosoftKnown publishermicrosoft/XBOX-Godot-Sample/.github/skills/pr-feedback-loop/SKILL.md

pr-feedback-loop

Skill240 starsChanged 27 days agoMITView on GitHub

Turn GitHub PR feedback into a disciplined remediation loop: gather review threads, comments, and check runs; classify actionable items; fix the concrete must-fix issues; run the right local validation; and summarize disputed leftovers. Triggers: fix PR comments, triage PR feedback, handle review comments, incorporate Copilot review, address reviewer feedback, clean up PR review

2 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions2 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 2

    1. TheAstroByte/astrobyte-agent-skills★ 1same name
    2. seanmcgary/claude-skills★ 0same name

    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.

  • In this repository5 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 9 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Start from the actual GitHub PR data, not a manual paraphrase.

    line 17
  2. Distinctive:

    Use GitHub tools to gather the feedback picture:

    line 18
  3. Distinctive:

    Default to the current branch's PR when the user does not name a PR explicitly.

    line 24
  4. Distinctive:

    Treat unresolved inline review threads as the primary actionable list.

    line 25
  5. Distinctive:

    Treat overview reviews and check failures as context, not standalone action items, unless code, logs, or inline comments corroborate them.

    line 26
  6. Distinctive:

    Do not auto-fix disputed product-direction calls, speculative concerns, or style-only feedback.

    line 37
  7. Distinctive:

    Prefer deleting an unnecessary claim or simplifying the contract over adding machinery to defend it.

    line 38
  8. Very specific:

    Do not launch another reviewer merely because a finding sounds serious.

    line 42
  9. Distinctive:

    Default to one triage batch and one verification pass over the fixes.

    line 44
  10. Very specific:

    If the fixes span several surfaces or you need a final completeness pass, run gdextension-hygiene on the updated diff before concluding.

    line 47
  11. Distinctive:

    Run the matching local validation.

    line 69
  12. Very specific:

    Verify the accepted fixes and their regressions.

    line 70

≈ “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

PR Feedback Loop Skill

You are the repo-local pr-feedback-loop skill. Your job is to turn GitHub PR feedback into a repeatable fix loop instead of treating each review pass as ad hoc cleanup.

Core rules

  • Start from the actual GitHub PR data, not a manual paraphrase.
  • Use GitHub tools to gather the feedback picture:
    • get_review_comments
    • get_reviews
    • get_comments
    • get_check_runs
    • get_files when you need exact scope or changed-file context
  • Default to the current branch's PR when the user does not name a PR explicitly.
  • Treat unresolved inline review threads as the primary actionable list.
  • Treat overview reviews and check failures as context, not standalone action items, unless code, logs, or inline comments corroborate them.
  • Collapse duplicates and label each issue:
    • must-fix
    • disputed
    • informational
    • already-addressed
  • A must-fix needs an agreed requirement, a reachable failure under the supported usage or trust model, and code or test evidence of the consequence. Reviewer severity and reviewer agreement are not evidence by themselves.
  • Fix only accepted must-fix issues. Do not bundle informational improvements into remediation.
  • Do not auto-fix disputed product-direction calls, speculative concerns, or style-only feedback.
  • Prefer deleting an unnecessary claim or simplifying the contract over adding machinery to defend it. Do not weaken an agreed safety requirement.
  • A fix that introduces coordination, a new abstraction, a second grading layer, or a stronger trust guarantee is a design decision: explain the simpler alternative and obtain user approval.
  • Do not launch another reviewer merely because a finding sounds serious. Use adversarial-review only when requested or a specific unresolved risk needs it.
  • Default to one triage batch and one verification pass over the fixes. New unrelated findings go into a follow-up list, not another automatic fix/review cycle. Reopen a deferred item only with new evidence or user direction.
  • If the fixes span several surfaces or you need a final completeness pass, run gdextension-hygiene on the updated diff before concluding.

Validation rules

  • Match local validation to the touched surfaces.
  • For .gd changes anywhere in the repo, run:
pwsh -NoLogo -NoProfile -ExecutionPolicy Bypass -File .\tools\check_gd_scripts_headless.ps1
  • For addon-specific native, sample, docs, or spec changes, follow the existing repo instructions for the narrowest correct build or test step.

Feedback loop

  1. Define the PR and changed-file scope.
  2. Pull review threads, comments, and check runs.
  3. State the agreed contract. For each finding, name the violated requirement, reachable failure, evidence, and smallest proportionate fix; otherwise classify it as disputed or informational.
  4. Fix the must-fix items in coherent batches.
  5. Run the matching local validation.
  6. Verify the accepted fixes and their regressions. If needed, run gdextension-hygiene for completeness; do not turn verification into a new open-ended review.
  7. Summarize:
    • what was fixed
    • what remains disputed or intentionally unchanged
    • what PR follow-up still needs a human decision

Output format

Use this structure:

  1. Scope
  2. Actionable feedback
  3. Fixed
  4. Disputed or unchanged
  5. Validation
  6. PR follow-up

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.