agentleFS
thoitiettxl-cyberthoitiettxl-cyber/Claw-II/skills/code-review-workflow/SKILL.md

code-review-workflow

Skill2 starsChanged 7 months agoMITView on GitHub

Consolidate the review workflow for OpenClaw-Module Android into a phase-aware skill with 3 modes: prepare review request, perform code review, and handle reviewer feedback; Prioritize choosing the correct mode according to the state of the review round instead of letting review-family skills compete for triggers. Use when request has clear review-loop intent: user wants to prepare packet review/handoff, wants assistant to review code/diff directly, or needs to process existing review comments.

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

10 rules. 0 are common, 9 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 9

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    No diff/change/comment context exists and the user is not asking about review workflow itself.

    line 57
  2. Distinctive:

    If review comments imply a real bug, hand off to debugging/verification after classifying the review phase.

    line 60
  3. Distinctive:

    If user wants external posting/writing, require approval before sending.

    line 61
  4. Distinctive:

    Do not package a review request when the user actually wants direct review.

    line 73
  5. Distinctive:

    Do not review code when the user only needs help answering comments.

    line 74
  6. Distinctive:

    Never overclaim safety, correctness, or reviewer validity without support.

    line 78
  7. Distinctive:

    Ask for the smallest missing input needed for the selected mode.

    line 84
  8. Distinctive:

    Must choose one primary mode.

    line 102
  9. Distinctive:

    Must not conflate reviewer-request packaging with direct review or comment response.

    line 103
  10. Distinctive:

    Must be evidence-backed and phase-appropriate.

    line 104

“Common” is measured: kept by at least 6 in 10 of the 3 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

Code Review Workflow

Indexed Reference Map

  • @[SKILL.md#L1-L99] Frontmatter, trigger rules, SOP, output contract.
  • @[references/INDEX.md#L1-L13] Reading order and coverage map.
  • @[references/review-modes.md#L1-L16] Mode selection and precedence.
  • @[references/review-request-template.md#L1-L14] Reviewer handoff artifact structure.
  • @[references/review-findings-rubric.md#L1-L20] Findings classes, evidence, and confidence rubric.
  • @[references/review-response-template.md#L1-L14] Comment-response structure and decision rules.
  • @[references/examples.md#L1-L5] Example prompt-to-mode mappings.

Scope

  • In-scope:
    • Preparing a reviewer-ready request after implementation.
    • Reviewing local diff, staged changes, or PR patch directly.
    • Interpreting and responding to review comments with evidence.
    • Selecting the correct review phase/mode before acting.
  • Out-of-scope:
    • Blindly mixing all review phases in one vague response.
    • Auto-posting comments to external services without approval.
    • Replacing debugging or final merge/branch closeout workflows.

KB Mode Notes

  • Reading order:
    1. references/INDEX.md
    2. references/review-modes.md
    3. references/review-request-template.md
    4. references/review-findings-rubric.md
    5. references/review-response-template.md
    6. references/examples.md
  • Coverage Matrix:
    • Mode selection -> review-modes.md
    • Request-review packaging -> review-request-template.md
    • Performing review findings -> review-findings-rubric.md
    • Reviewer response handling -> review-response-template.md
    • Prompt examples -> examples.md
  • Portability note:
    • Workflow is review-methodology-first and local-first; transport/provider specifics stay out of scope.

Trigger Rules

  • Positive:
    • User wants to prepare a review request, review packet, reviewer handoff, or pre-PR summary.
    • User explicitly asks the assistant to review code, diff, patch, or PR directly.
    • User already has reviewer comments and asks how to respond, whether a comment is valid, or what should be patched.
    • Prompt contains mixed review intent and the assistant must choose the correct phase first.
  • Negative:
    • Request is mainly planning, debugging, implementation, or final branch closeout with no real review-loop intent.
    • User only wants formatting/lint cleanup with no review workflow.
    • No diff/change/comment context exists and the user is not asking about review workflow itself.
    • The user only wants generic advice about code quality/process, not an actual review-phase action.
  • Escalation:
    • If review comments imply a real bug, hand off to debugging/verification after classifying the review phase.
    • If user wants external posting/writing, require approval before sending.

SOP

  1. Select mode first
    • Choose one primary mode: REQUEST_REVIEW, PERFORM_REVIEW, or RESPOND_TO_REVIEW.
  2. State the mode briefly when useful
    • Especially if the prompt could have matched multiple old skills.
  3. Run the phase-specific workflow
    • Request mode -> package scope, summary, evidence, and focus areas.
    • Perform mode -> inspect diff/context, report actionable findings with confidence.
    • Respond mode -> classify comment, verify evidence, then accept/clarify/push back/test-first.
  4. Avoid phase bleed
    • Do not package a review request when the user actually wants direct review.
    • Do not review code when the user only needs help answering comments.
  5. Mention next phase only if natural
    • e.g. after request packet, next likely phase is actual review.
  6. Keep evidence discipline
    • Never overclaim safety, correctness, or reviewer validity without support.
  7. Escalate appropriately
    • Debugging/verification can be follow-up skills when evidence demands it.

Error Handling

  • Missing diff/comment/context:
    • Ask for the smallest missing input needed for the selected mode.
  • Mixed review stages in one request:
    • Pick the earliest blocking stage first and name the likely next stage.
  • Weak evidence:
    • Downgrade confidence or return PARTIAL instead of overstating certainty.
  • User requests blind agreement or blind review posting:
    • Refuse and explain the missing verification/approval gate.

Output Contract

  • Deliverable:
    • Mode: REQUEST_REVIEW | PERFORM_REVIEW | RESPOND_TO_REVIEW
    • Goal: what this review-phase output is trying to achieve.
    • Mode-specific body:
      • Request -> scope, summary, evidence, focus areas, open questions, suggested review request.
      • Perform -> summary, findings, confidence, next actions.
      • Respond -> comment, classification, assessment, evidence, recommended response, next step, confidence.
    • Possible Next Phase: only if obvious.
  • Validation:
    • Must choose one primary mode.
    • Must not conflate reviewer-request packaging with direct review or comment response.
    • Must be evidence-backed and phase-appropriate.
  • Compliance Verdict:
    • PASS when the chosen mode and output fit the review state; otherwise PARTIAL or FAIL with reason.

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.