RE

review-quality

Formalizes the review process by verifying plan alignment, test quality, and code simplicity before finalizing implementation.

Install

mkdir -p .claude/skills/review-quality && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/12208" && unzip -o skill.zip -d .claude/skills/review-quality && rm skill.zip

Installs to .claude/skills/review-quality

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Review quality framework for the work-to-review transition gate. Guides verification of plan alignment, test quality, and code simplification before marking implementation complete. Referenced by schema guidance fields during review-phase note filling. Use when filling review-checklist notes or when asked to review completed implementation work.
347 chars✓ has a “when” triggerlonger than Claude Code's old 250-char listing cap (fine on current versions)
Advanced

Key capabilities

  • →Load MCP item notes for review
  • →Read changed files identified in implementation notes
  • →Run and capture results of the project's test suite
  • →Compare implemented work against planning notes
  • →Verify test coverage against the planning note's test strategy
  • →Evaluate test substance for meaningful assertions

How it works

This framework guides a reviewer to verify implementation work by loading item notes, reading changed files, running tests, and comparing the work against planning and test strategies.

Inputs & outputs

You give it
MCP item ID, planning note, implementation notes, changed files, test suite
You get back
Review-checklist note with a verdict (Pass, Fail, Pass with observations) and findings

When to use review-quality

  • →Review implementation against feature plans
  • →Verify test coverage and quality
  • →Check for unnecessary code complexity
  • →Validate bug fix implementation

About this skill

Review Supplement for This Codebase

This supplements the generic review method; it never repeats it.

Build state

Read the build state the orchestrator recorded (the review scoping slot's build-state line, the session-tracking Verified: line or the test-manifest VERIFY line) and run no Gradle suite: rule test-author section 5 points 8 and 11. A missing build state, or one for an older SHA than the diff under review, is a blocking finding.

Test substance (every reviewed item, with or without needs-test-author)

Fetch the served rules forbidden-test-patterns and test-assertion-vacuity as rules (config op rules, key, rootId) and judge every test the diff adds or changes against them. Also check for happy-path-only coverage: tests that cover only the happy path of a change that has failure and edge cases are a gap to report.

Items with the needs-test-author trait

  • Inputs: the test-plan and test-manifest notes, and the orchestrator's per-child SHA table (Pre-SHA, Post-SHA, Test-Pre-SHA, Test-Post-SHA) with any declared fixture-repair SHAs. A trait-bearing item with no test-plan note is a blocking issue on its own.
  • Declared orchestrator fixture repairs are judged by rule test-file-ownership point 4 and rule review-scoping point 3; they may touch only construction or setup code, never assertions.
  • An oracle that matches the implementation but not its cited source is blocking.
  • A<n> lines in orchestrator-confirmation are red-test rulings (rule arbitration).
  • A false forbidden-pattern declaration weighs as not-independent.
  • An independence result of independent-degraded (temporal-only) maps to Pass: record it as the observed mode, not as a finding.
  • Spot-check the coverage mapping: open at least two mapped tests and read their bodies rather than trusting the manifest alone.
  • Check the planning freeze: when the schema declares amendments, the planning note's modifiedAt (notes_read list) is its freeze time; otherwise its ## Amendment A<n> sections are the only later edits. Compare the freeze with the implementer's first commit.

Items with the needs-docs-accuracy-review trait

Rule review-cited-claims point 1 governs cited lines; on a docs item it also covers claims of defaults, limits, budgets and ordering, and wording copied from an existing doc is unverified until its source is opened.

  • Open each claim -> source file:line pair the planner lists.
  • Check the Docs and CHANGELOG lines against git diff --name-only <base>...HEAD.
  • Scan changed files for bytes below 0x20 other than TAB and LF, non-UTF-8 bytes and CR line endings.

Two checks the generic method lacks

  • Root cause. A fix addresses the diagnosed root cause, not its symptom.
  • Scenario coverage. Every scenario in the test strategy and every edge case in the diagnosis maps to a test; a missing one is a gap to report.

When not to use it

  • →When the reviewer is the same agent that wrote the code or tests
  • →When planning notes or implementation notes are missing

Limitations

  • →The reviewer must not be the same agent that wrote the code or the tests.
  • →If the planning note or implementation notes are missing, the review cannot proceed.
  • →The reviewer does not advance the item.

How it compares

This skill provides a structured, independent review process to catch common failure modes, unlike an unguided review or self-assessment.

Compared to similar skills

review-quality side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
review-quality (this skill)04moNo flagsAdvanced
python-testing-patterns776moReviewIntermediate
dependency-upgrade268moReviewIntermediate
test-cases5710moNo flagsBeginner

Try saying

Example prompts that trigger this skill in your AI assistant.

Search skills

Search the agent skills registry