| name | requesting-code-review |
| description | Use when completing tasks, implementing major features, or before merging to verify work meets requirements. Dispatches the repo-local code-reviewer subagent. |
Requesting Code Review
Dispatch the repo-local code-reviewer subagent to catch issues before they cascade. The reviewer gets precisely crafted context for evaluation — never your session's history. This keeps the reviewer focused on the work product, not your thought process, and preserves your own context for continued work.
Core principle: Review early, review often.
Where the reviewer lives
The code-reviewer subagent definition is the single source of truth for the reviewer's checklist, severity rules, and output format. In this template it lives at agents/code-reviewer.md. The scaffold installs it into the downstream repo at:
.claude/agents/code-reviewer.md for Claude Code sessions.codex/agents/code-reviewer.md for Codex sessions (only when Codex is set up)
The .claude/ copy is always installed by the template scaffold; the .codex/ copy is added only when Codex is enabled. If the Claude copy is missing, run update-skills against the repo to install it.
The prompt template at brief.md (alongside this SKILL.md) is only the dynamic brief — placeholders for what was built, the plan, and the SHA range. It deliberately does not repeat the checklist; the agent definition already has it.
When to Request Review
Mandatory:
- After each task in subagent-driven development
- After completing a major feature
- Before merge to main
Optional but valuable:
- When stuck (fresh perspective)
- Before refactoring (baseline check)
- After fixing a complex bug
How to Request
1. Get git SHAs:
BASE_SHA=$(git rev-parse HEAD~1)
HEAD_SHA=$(git rev-parse HEAD)
2. Dispatch the code-reviewer subagent:
- Claude session: call the Task tool with
subagent_type: code-reviewer and a prompt built from the template at requesting-code-review/brief.md (this file is shipped alongside the SKILL.md).
- Codex session: invoke the equivalent subagent dispatch with the same prompt template.
Placeholders to fill in the prompt:
{WHAT_WAS_IMPLEMENTED} — what you just built{PLAN_OR_REQUIREMENTS} — what it should do (cite the plan file or bead description){BASE_SHA} — starting commit{HEAD_SHA} — ending commit{DESCRIPTION} — brief summary
The subagent will run git diff itself, read the changed files, and return a structured review (Strengths / Critical / Important / Minor / Assessment). It is read-only — it cannot modify code.
The reviewer always grounds itself in the repo's layered CLAUDE.md conventions (defined in the agent definition, noted here so callers know to expect it): for every changed file it reads the nearest CLAUDE.md walking up from the file's directory (e.g. apps/web/CLAUDE.md for web changes, packages/utils/CLAUDE.md for utils changes) plus the root CLAUDE.md, and judges the diff against those rules — including the root comment-discipline convention — not generic best practices.
3. Act on feedback:
- Fix Critical issues immediately
- Fix Important issues before proceeding
- Note Minor issues for later
- Push back if the reviewer is wrong (with reasoning)
Example
[Just completed Task 2: Add verification function]
You: Let me request a code review before proceeding.
BASE_SHA=$(git log --oneline | grep "Task 1" | head -1 | awk '{print $1}')
HEAD_SHA=$(git rev-parse HEAD)
[Dispatch code-reviewer subagent]
WHAT_WAS_IMPLEMENTED: Verification and repair functions for conversation index
PLAN_OR_REQUIREMENTS: Task 2 from docs/plans/2026-04-30-deployment.md
BASE_SHA: a7981ec
HEAD_SHA: 3df7661
DESCRIPTION: Added verifyIndex() and repairIndex() with 4 issue types
[Subagent returns]:
Strengths: Clean architecture, real tests
Issues:
Important: Missing progress indicators
Minor: Magic number (100) for reporting interval
Assessment: Ready to proceed (with fixes)
You: [Fix progress indicators]
[Continue to Task 3]
Integration with Workflows
Subagent-driven development:
- Review after EACH task
- Catch issues before they compound
- Fix before moving to next task
Executing plans:
- Review after each batch (3 tasks)
- Get feedback, apply, continue
Ad-hoc development:
- Review before merge
- Review when stuck
Red Flags
Never:
- Skip review because "it's simple"
- Ignore Critical issues
- Proceed with unfixed Important issues
- Argue with valid technical feedback
If reviewer wrong:
- Push back with technical reasoning
- Show code/tests that prove it works
- Request clarification
Files in this skill
SKILL.md — this filebrief.md — the prompt template (dynamic brief only) you fill in and pass to the subagent
The reviewer's standing instructions live separately at agents/code-reviewer.md in the template root and are deployed by the scaffold to .claude/agents/code-reviewer.md (and .codex/agents/code-reviewer.md when Codex is set up).
Pairing with verification-before-completion
The executor chain in executor-task and executor-task-worktree runs verification-before-completion first (run the verification commands and read the output), then requesting-code-review (dispatch the reviewer subagent). Both are required before beads-close — they are NOT alternatives. verification-before-completion proves the change passes its own checks; requesting-code-review gives an independent reviewer's perspective on quality, regressions, and architectural drift before the bead closes.
