Agent-Skills.md

Agent Skills: Code Review Skill

Review implementation for code quality, style consistency, and best practices. Runs before security review. READ-ONLY - reports issues but does not fix them.

UncategorizedID: matthew-plusprogramming/monorepo/code-review

Repository

matthew-plusprogramming/monorepo
matthew-plusprogramming

Install this agent skill to your local

pnpm dlx add-skill https://github.com/matthew-plusprogramming/monorepo/tree/HEAD/.claude/skills/code-review

Skill Files

Browse the full folder contents for code-review.

Download Skill

Loading file tree…

.claude/skills/code-review/SKILL.md

Skill Metadata

Name
code-review
Description
Review implementation for code quality with style/naming, test-quality, adversarial, and holistic passes. Runs before security review. READ-ONLY - reports issues but does not fix them.

Code Review Skill

Purpose

Review the completed implementation for quality issues before security review. The review is read-only and tied to one approved spec.md.

Key input: .claude/specs/groups/<spec-group-id>/spec.md

Usage

/code-review <spec-group-id>

Prerequisites

Before dispatch:

  1. Spec group exists.
  2. /unify has passed or produced a reviewable partial result.
  3. Changed files and validation evidence are available.
  4. Tests relevant to the spec have run or failures are documented.

If prerequisites are missing, stop and run /unify or complete the missing validation first.

Required Review Specialties

Every code-review report must include these four sections:

| review_specialty | Focus | | ------------------ | ----- | | style_naming | Redundancy, conventions, DRY, naming, API shape, local maintainability | | test_quality | Assertion strength, vacuous truth, tautology, weak snapshots, isolation, determinism | | adversarial | How this could pass incorrectly, false positives, happy-path bias, unproven invariants | | holistic | Whole-change coherence, duplicate consolidation, severity normalization, final judgment |

Every finding must include review_specialty.

Process

  1. Load manifest.json, spec.md, unifier output, changed-file list, and reviewer-focus metadata when present.
  2. If a dispatch retry needs prior context, read .claude/coordination/review-dispatch-prompt-<dispatch-id>.json.
  3. Build review focus from reviewer-focus metadata, the spec's risk areas, contracts, edge cases, and validation evidence.
  4. Review changed files against the four specialties.
  5. Check tests for meaningful AC coverage and failure modes.
  6. Normalize duplicate findings and severity.
  7. Return pass/fail recommendation with concrete file references.

Finding Format

Each finding should include:

  • severity: critical, high, medium, or low
  • review_specialty
  • file
  • line
  • issue
  • impact
  • recommendation

Output

Return:

  • status: pass, fail, or partial
  • finding_count_by_severity
  • top_blockers
  • findings
  • residual_risk

After a clean pass, update convergence:

node .claude/scripts/session-checkpoint.mjs update-convergence code_review