| name | code-review |
| description | Review git changes with project-specific risk checks layered on top of general senior code review. Use when asked for a code review, merge readiness assessment, or to flag likely regressions. |
Code Review
Perform a project-specific review of current changes.
When to Use
- The user asks for a code review.
- The user wants a merge/readiness assessment.
- The user needs likely regressions flagged.
Steps
-
Scope the diff: git status -sb, git diff --stat, git diff.
-
Classify touched areas and apply risk checks:
| Area | What to Check |
|---|
packages/adapters/ | preflight/execute contract, capability metadata, registry entry, no duplicate IDs, no secret leakage |
packages/judges/ | all judges return critical with a boolean default, regex safety, new type in TaskJudge union |
packages/report/ | weight presets sum to 1.0, getDefaultWeights is the single source of truth, no duplicate definitions |
packages/runner/ | imports weights from report package (not local copy), no duplicate judge execution, scoreWeights written to result |
packages/cli/ | --score-mode valid modes match scoring.ts exactly, new params passed through to runner |
packages/taskpacks/ | new enum fields validated, new judge types parsed, supportedTypes array updated |
packages/core/ | new interfaces exported, union types updated, backward compatibility (new fields optional) |
apps/web-report/ | weights match backend, default mode = "practical", i18n in both languages, design tokens used |
tests/ | new features have tests, no tests silently skipped |
-
Run automated validation: pnpm build && pnpm test.
-
Output findings ordered by severity, with file/line references.
Stop Conditions
- The review gives generic feedback without checking project-specific risk areas.
- A scoring weight change was reviewed without verifying the sum equals 1.0.
- An adapter or judge change was reviewed without running build + tests.
