Dispatch superpower-planning:code-reviewer subagent to catch issues before they cascade.
Core principle: Review early, review often.
When to Request Review
Mandatory:
- After each task in subagent-driven development
- After completing a major project step
- Before merge to main
- When implementation may have drifted from the approved plan
Optional but valuable:
- When stuck (fresh perspective)
- Before refactoring (baseline check)
- After fixing a complex bug
- After making architecture-affecting changes
Review Priorities
Every requested review should explicitly cover:
- Plan alignment — does the implementation match the approved plan/spec?
- Code quality — error handling, maintainability, naming, structure, tests
- Architecture — boundaries, coupling, integration, unnecessary complexity
- Documentation and standards — comments/docs/conventions where relevant
- Severity semantics — Critical / Important / Minor
Standard Dispatch Template
Use the detailed template at skills/requesting-review/code-reviewer.md, but gather the following inputs first.
Required Inputs
{WHAT_WAS_IMPLEMENTED}— what was just built{PLAN_OR_REQUIREMENTS}— task text, plan section, or approved requirements{BASE_SHA}— start of the diff range{HEAD_SHA}— end of the diff range{DESCRIPTION}— brief summary of the change set
Recommended Inputs
- plan file path
- spec file path
- task number / task heading
- whether this is an intermediate task review or final implementation review
How to Request
1. Get git SHAs
BASE_SHA=$(git rev-parse HEAD~1) # or origin/main / last reviewed commit
HEAD_SHA=$(git rev-parse HEAD)
2. Dispatch code-reviewer subagent
Use Task tool with superpower-planning:code-reviewer and fill the template at skills/requesting-review/code-reviewer.md.
3. Minimum Dispatch Shape
WHAT_WAS_IMPLEMENTED: [brief statement]
PLAN_OR_REQUIREMENTS: [exact task text / plan section / requirements]
BASE_SHA: [sha]
HEAD_SHA: [sha]
DESCRIPTION: [brief summary]
4. Preferred Dispatch Shape
WHAT_WAS_IMPLEMENTED: Task 3 - verification and repair functions
PLAN_OR_REQUIREMENTS: Task 3 from docs/plans/deployment-plan.md
PLAN_PATH: docs/plans/deployment-plan.md
SPEC_PATH: docs/plans/deployment-design.md
BASE_SHA: a7981ec
HEAD_SHA: 3df7661
DESCRIPTION: Added verifyIndex() and repairIndex() with 4 issue types
REVIEW_MODE: intermediate task review
FOCUS: plan alignment, tests, error handling
How to Interpret Feedback
Critical
- Must fix before proceeding or merging
- Includes broken functionality, data loss risk, security issues, severe requirement misses
Important
- Should fix before proceeding when practical
- Includes architecture gaps, weak tests, bad error handling, missing requirement edges
Minor
- Nice to have
- Includes readability, polish, documentation cleanup, non-blocking refactors
Acting on Feedback
- Fix Critical issues immediately
- Fix Important issues before proceeding unless you have a strong technical reason not to
- Note Minor issues for later if time-sensitive
- If reviewer is wrong, push back with reasoning and evidence
- If reviewer identifies plan drift, either:
- fix implementation to match the plan, or
- explicitly update the plan/spec if the deviation is justified
After Review Results
- Append review outcome (approved / changes-requested, issues fixed, issues deferred) to
.planning/progress.mdsession log - Append technical insights (patterns learned, architectural feedback worth remembering) to
.planning/findings.mdunder## Code Review Findings - If the review surfaced plan/spec drift, record that explicitly in
.planning/findings.md
Example
[Just completed Task 2: Add verification function]
You: Let me request code review before proceeding.
BASE_SHA=$(git log --oneline | grep "Task 1" | head -1 | awk '{print $1}')
HEAD_SHA=$(git rev-parse HEAD)
[Dispatch superpower-planning:code-reviewer subagent]
WHAT_WAS_IMPLEMENTED: Verification and repair functions for conversation index
PLAN_OR_REQUIREMENTS: Task 2 from docs/plans/deployment-plan.md
PLAN_PATH: docs/plans/deployment-plan.md
SPEC_PATH: docs/plans/deployment-design.md
BASE_SHA: a7981ec
HEAD_SHA: 3df7661
DESCRIPTION: Added verifyIndex() and repairIndex() with 4 issue types
REVIEW_MODE: intermediate task review
FOCUS: plan alignment, tests, error handling
[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:
- Review after each task
- Use review output as a hard quality gate before marking task complete
Executing Plans:
- Review after each batch or major milestone
- Apply fixes before continuing
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 casually
- Ask for review without giving plan/requirements context
- Treat code review as style-only; plan alignment matters too
If reviewer seems wrong:
- Push back with technical reasoning
- Show code/tests that prove it works
- Request clarification
- Update plan/spec if the implementation is intentionally better than the original plan
See template at: skills/requesting-review/code-reviewer.md