bitranox Requesting Code Review
Adapted from the superpowers plugin (MIT).
Dispatch a 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.
When to Request Review
Mandatory:
- After each task in subagent-driven development
- After completing major feature
- Before merge to main
Optional but valuable:
- When stuck (fresh perspective)
- Before refactoring (baseline check)
- After fixing complex bug
How to Request
1. Get git SHAs:
BASE_SHA=$(git rev-parse HEAD~1) # or origin/main
HEAD_SHA=$(git rev-parse HEAD)
2. Dispatch code reviewer subagent:
Dispatch a general-purpose subagent, filling the template at code-reviewer.md
Placeholders:
[DESCRIPTION]- Brief summary of what you built[PLAN_OR_REQUIREMENTS]- What it should do[BASE_SHA]- Starting commit[HEAD_SHA]- Ending commit
3. Act on feedback:
- Fix Critical issues immediately
- Fix Important issues before proceeding
- Note Minor issues for later
- Push back if reviewer is wrong (with reasoning)
Before merging: scan the push range for leaked secrets, private infrastructure, or personal data (see bitranox:compuse-git) - a reviewer should flag these too.
A gate is real only when the actor cannot satisfy it
A review is a gate, and every gate fails the same way: the actor whose work is checked also decides whether the check passed. A gate is real only when that actor is STRUCTURALLY incapable of satisfying it - not merely instructed not to. Ask what the actor would do to pass WITHOUT doing the work; if the answer is "assert that it did", the gate is advice.
| Does not gate | Gates, because the actor cannot reach it |
|---|---|
| a role name in a prompt ("act as a reviewer") | a state machine checked against an authenticated identity |
| an LLM grading a peer, or itself | a separate job with its own permissions |
| a regex over prose the agent writes | a non-AI step returning an exit code |
An LLM reviewer FINDS; it does not AUTHORIZE. Both are true at once: dispatching the reviewer is worth doing - fresh context and no sunk cost let it see what you cannot - and it is still not the gate. Keep the jobs apart. The review surfaces candidates; something with an exit code decides whether the work proceeds. Separateness buys better FINDING, never gating, so reading a subagent's verdict as permission is exactly what puts an LLM in the left-hand column.
This is why the reviewer here is a SEPARATE subagent that never saw your session, and why "I already tested it" is a Red Flag below. Apply it to the findings too - but gate on the FIX, not the FINDING: checking prose for a finding is the left-hand column again, whereas a reported fix is a claim until a command exits non-zero on the old code and zero on the new.
Measured across five agent orchestrators: four shipped verification that could not refuse, and what separated the fifth was purely WHERE the gate lived - not effort, not code quality.
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 code reviewer subagent]
DESCRIPTION: Added verifyIndex() and repairIndex() with 4 issue types
PLAN_OR_REQUIREMENTS: Task 2 from docs/plans/deployment-plan.md
BASE_SHA: a7981ec
HEAD_SHA: 3df7661
[Subagent returns]:
Strengths: Clean architecture, real tests
Issues:
Important: Missing progress indicators
Minor: Magic number (100) for reporting interval
Assessment: Ready to proceed
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 task or at natural checkpoints
- 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
These thoughts mean STOP - you are rationalizing:
| Excuse | Reality |
|---|---|
| "It's simple, it doesn't need review" | Simple diffs ship the confident bugs. Size is not risk. |
| "I already tested it" | Tests check what you thought of; review catches what you didn't. |
| "The reviewer will just nitpick" | Then the nitpicks are cheap. Critical findings are not. |
| "I'll request review after one more fix" | The fix is exactly what needs reviewing. |
| "I wrote it, I know it's right" | Authorship is the reason you cannot see it. |
If reviewer wrong:
- Push back with technical reasoning
- Show code/tests that prove it works
- Request clarification
See template at: code-reviewer.md