Imported from lacucaracha421/chatgpt (.claude/skills/requesting-code-review/SKILL.md). Install upstream with npx skills add lacucaracha421/chatgpt --skill requesting-code-review. Copyright stays with the author.

Requesting Code Review

Use a supported reviewer when independent review adds value and delegation is available; otherwise perform the same read-only review inline and state that limitation. Provide requirements, scope, and evidence rather than the entire session history.

Core principle: Review early, review often.

When to Request Review

Request review when the user asks, an agreed acceptance gate requires it, or a substantial change warrants an independent perspective. Task completion alone does not mandate another reviewer or test run.

Optional but valuable:

When stuck (fresh perspective)
Before refactoring (baseline check)
After fixing complex bug

How to Request

1. Establish the complete review scope:

Follow Review scope. Record the verified base and head, task paths, committed/staged/unstaged layers, relevant untracked files, concurrent-work exclusions, and evidence. Do not default to HEAD~1; a feature branch's own upstream is not its task base. Inspect each layer as well as the combined final state, and disclose any unresolved coverage.

2. Review with a supported agent or inline fallback:

Use an available read-only reviewer or general-purpose subagent with code-reviewer.md. If no supported delegation exists, use that template inline; do not install a named reviewer/plugin.

Before treating the fallback as sufficient, check whether the user, task specification, applicable repository rule, or acceptance criteria explicitly require independent review:

Independent review is desirable, not required: perform useful scoped inline/self-review, disclose that it was not independent, and continue ordinary authorized work under the existing verification policy. Unavailable delegation alone does not block completion.
Independent review is an explicit completion/approval condition: perform useful scoped self-review and continue unrelated authorized implementation, but disclose that the review was not independent and keep the independent-review acceptance condition marked incomplete. Self-review cannot satisfy that condition; do not claim the gated completion/approval until the required independent review has occurred.

Do not install tools, expand permissions, invent a reviewer, or mislabel self-review as independent merely to satisfy the condition.

Placeholders:

{DESCRIPTION} - Brief summary of what you built
{PLAN_OR_REQUIREMENTS} - What it should do
{BASE_SHA} - Starting commit
{HEAD_SHA} - Verified ending commit
{TASK_PATHS} - Task-owned paths and relevant dependencies
{WORKTREE_SCOPE} - Whether staged/unstaged changes are included
{UNTRACKED_FILES} - Relevant untracked files to read directly
{EXCLUSIONS} - Out-of-scope/pre-existing/concurrent work
{EVIDENCE} - Inspected results and outstanding native/device gates

3. Act on feedback:

In read-only review, report findings without editing
In an authorized repair task, fix in-scope Critical/Important issues; pause only affected work if a decision or permission is missing
Note Minor issues for later
Push back if reviewer is wrong (with reasoning)

Example

[Just completed Task 2: Add verification function]

You: Let me request code review before proceeding.

BASE_SHA=<verified-task-start-commit>
HEAD_SHA=$(git rev-parse HEAD)

[Dispatch code reviewer subagent]
  DESCRIPTION: Added verifyIndex() and repairIndex() with 4 issue types
  PLAN_OR_REQUIREMENTS: Approved task requirements or the current plan path
  BASE_SHA: a7981ec
  HEAD_SHA: 3df7661
  TASK_PATHS: <task-owned paths>
  WORKTREE_SCOPE: include staged and unstaged task changes
  UNTRACKED_FILES: <relevant new files, or none>
  EXCLUSIONS: <unrelated concurrent changes, or none>
  EVIDENCE: <valid checks and remaining acceptance gates>

[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]

Common Rationalizations

Excuse	Reality
"Delegation is unavailable"	Perform scoped self-review and disclose it. Continue unrelated authorized work, but keep any explicitly mandatory independent-review condition incomplete. Do not install tools or expand permissions to satisfy it.
"The reviewer needs my whole session history to understand the change"	Hand it precisely crafted context, never your session's history. That keeps the reviewer on the work product, not your thought process.

Red Flags

Never:

Skip requested or risk-required review without disclosing the gap
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

See template at: code-reviewer.md

requesting-code-review

About