agentleFS
drosamdrosam/agents/skills/pr-review/SKILL.md

pr-review

Skill1 starsChanged 2 days agoMITView on GitHub

Prepares concise, high-signal GitHub reviews when someone requests the user's PR review. Use for "review this PR for me", "draft feedback for GitHub", "turn this code review into PR comments", or walking through outgoing review feedback. Runs code-review or reuses a selected review, refines comments with the user, recommends approve/request changes/comment, and publishes only after explicit approval. Not for exhaustive local audits or addressing incoming comments on the user's code (use review-triage).

195 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use oss-pr, pr-automation and 10 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.
  • CES-LtdSkill

    oss-pr

    Use when creating a pull request, after committing changes, or when user invokes /oss-pr. Covers branch management, quality checks, commit, push, and PR creation.

    Also in 3 projects that use this

  • CES-LtdSkill

    pr-automation

    PR Automation Orchestrator: poll open PRs, check CI, run review, fix, and merge eligible PRs. Use when: (1) Invoked by daemon via scripts/pr-automation.sh, (2) User says "/pr-automation".

    Also in 3 projects that use this

  • jtenniswoodSkill

    expert-reset

    Reimagine this repository from first principles using the current product and user experience as the target. Use when the user says "/expert-reset", asks how the project should be rebuilt if starting again, wants bold architecture or component-structure recommendations, or wants a research-only reset plan unconstrained by current implementation details.

    Also in 3 projects that use this

  • jtenniswoodSkill

    updates

    Summarize user-facing project updates from the last 7 days, ordered by importance and written as one-sentence "Feature: description" entries. Use when the user says "/updates", asks what changed recently, or wants a concise list of recent features, device support, fixes that users will notice, or other product value.

    Also in 3 projects that use this

  • openchamberSkill

    clack-cli-patterns

    Use when creating or modifying OpenChamber CLI commands, prompts, terminal output, non-TTY behavior, `--quiet`, or `--json` behavior.

    Also in 3 projects that use this

  • openchamberSkill

    drag-to-reorder

    Use when implementing or modifying OpenChamber sortable or drag-to-reorder behavior, especially `@dnd-kit`, touch/mobile interactions, variable-width items, or wrapping layouts.

    Also in 3 projects that use this

  • +6 more from openchamber

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 2 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 179 edited versions we read keep these.

  • 2

    Distinctive

    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. Very specific:

    Keep/Reword/Omit/Investigate decisions affect public feedback only.

    line 14
  2. Distinctive:

    Ask with the question tool: assessment plus draft in the Markdown question, then Keep / Reword / Omit / Investigate as applicable, …

    line 27
  3. Very specific:

    Read references/calibration.md when importance, disputed reachability or wording needs calibration.

    line 30
  4. Distinctive:

    Ask for explicit publication authorization of that exact version.

    line 41
  5. Very specific:

    Before posting, read references/github-publication.md.

    line 42

“Common” is measured: kept by at least 6 in 10 of the 179 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 Review

Investigate thoroughly, publish selectively. Flow: source review -> shortlist -> item walk -> event -> exact preview -> explicit publication approval.

Boundaries

  • Suggest fixes only: prose or complete anchored GitHub suggestion blocks. Never edit code, commit, push, apply suggestions, resolve threads, dismiss reviews or merge. Implementing a fix requires leaving this workflow.
  • Draft locally. Creating even a pending remote review is a write that needs publication authorization.
  • Keep/Reword/Omit/Investigate decisions affect public feedback only. Omitting a finding does not resolve it; serious unresolved risks still constrain the event recommendation.

1. Source review

  1. Resolve repository, canonical PR URL/number, base/head SHAs, head branch, state and existing discussion with read-only gh. A closed/merged PR may be reviewed historically but never submitted as an active review.
  2. Fresh PR: load code-review and run it normally; it persists review.md and returns Artifact: <path>. Existing report: use only an explicitly selected path; read it whole, do not rerun discovery, but check current head/discussion against its revision and re-review changed paths.
  3. Create append-only feedback.md beside the selected review.md. Use the same log conventions as review-triage's triage.md (identity header, Source: ./review.md, one dense timestamped line per event, exact decisive user input as a JSON string), with the outgoing vocabulary Keep/Reword/Omit/Investigate plus draft versions, event choice and publication outcome. If the file already exists with conflicting identity, ask. review.md stays immutable; corrections and revalidation go in feedback.md under the original finding IDs.

2. Shortlist, then walk items

  • Propose which findings to publish and which to omit, one short reason each, keeping stable IDs and source order. Recommendations are not decisions; no silent omission, merge or reorder.
  • Prioritize introduced/worsened defects with real impact, broken contracts and substantial maintenance cost. Recommend omitting taste, cleanup nits, feedback already in a thread, disproven claims and unrelated pre-existing problems. Do not filter by inherited P-labels, round votes or a comment quota.
  • Investigate only the current item; verify its trigger, impact and counterevidence independently of the report's conclusion. Keep impact, reachability (UI vs API vs crafted request) and unknown frequency separate; a race is not rare without evidence. If evidence is incomplete, name the settling check and offer Investigate or an explicitly uncertain question draft.
  • Ask with the question tool: assessment plus draft in the Markdown question, then Keep / Reword / Omit / Investigate as applicable, recommendation first, single-select. Keep means include this wording, not publish. Reword/Investigate keeps the item open.
  • Log the decision, confirm briefly, pause for continuation before the next item unless the user explicitly waived stops for this session. Batch authority does not waive stops.

Read references/calibration.md when importance, disputed reachability or wording needs calibration.

3. Draft style

Write every draft with writing-for-humans; the rules here win where they differ. Natural short sentences, usually 1-3 per comment. Lead with what goes wrong and when, then concrete impact; add an example only when it clarifies the mechanism. Suggest the smallest sound fix; use a suggestion block only when replacing that exact range fully fixes the issue and syntax/companion changes are verified. No finding IDs, Problem/Evidence/Fix labels, severity boilerplate, filler praise or repeating inline comments in the body. Mark nonblocking suggestions as such. Evidence detail stays local unless the author needs it to act.

4. Event, preview, publication

  1. Recommend one event with a short reason, independent of the internal BLOCKED/INCOMPLETE/READY label: APPROVE (no unresolved blockers), REQUEST_CHANGES (verified material problems), COMMENT (useful feedback or uncertainty prevents approval). Missing runtime checks alone are not a defect.
  2. User chooses; log recommendation and choice separately. If the choice conflicts with known blockers, say so and clarify.
  3. Show the complete exact payload: body, inline comments with anchors, suggestion blocks, event, PR URL and head SHA, as a numbered draft version. Any wording change needs renewed acceptance.
  4. Ask for explicit publication authorization of that exact version. "Looks good", choosing APPROVE or the original review request is not authorization. Do not re-ask when authorization is explicit and unchanged.
  5. Before posting, read references/github-publication.md. Changed head, content, event or relevant discussion invalidates approval: re-preview and re-ask. Log the result with URLs; an ambiguous response is blocked, not posted, and logging failure blocks any retry until remote state is reconciled.

Return short progress summaries with artifact paths; the full investigation stays in review.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.