搜索

PythonReactSEOPDF
GitHub Skill

review-criteria

Criteria for reviewing Codex worker outputs across 5 dimensions. Use when reviewing code submitted by workers before approval.

DUTSTECH☆ 0 仓库 Stars0 仓库 Forks

安装与使用

请根据来源仓库和你的 AI 工具说明安装此技能。先阅读技能文件和依赖要求,再添加到工作流。

查看来源 ↗使用指南

技能说明

Review Criteria Skill

This skill defines how the CEO (Claude) reviews outputs from Codex workers.

Review Framework

Every Codex output must be reviewed across 5 dimensions:

1. Completeness (Required)

Question: Did the worker do what was asked?

Checklist:

  • All files in task specification were created/modified
  • The "Do" instruction was followed
  • Output contains completion signal (TASK_COMPLETE)

Fail Conditions:

  • Missing files that were specified
  • Partial implementation with TODO comments
  • No completion signal
2. Acceptance Criteria (Required)

Question: Does the output meet all acceptance criteria?

Process:

For each acceptance criterion in the task:

- [ ] AC-1: {criterion} → PASS/FAIL: {reason}
- [ ] AC-2: {criterion} → PASS/FAIL: {reason}
- [ ] AC-3: {criterion} → PASS/FAIL: {reason}

Criteria Met: {passed}/{total}

Fail Conditions:

  • Any P0/P1 criterion fails
  • More than 50% of criteria fail
3. Code Quality (Required)

Question: Is the code well-written?

Aspect Weight Check
Patterns High Follows existing codebase patterns
Bugs High No obvious logical errors
Error Handling Medium Proper try/catch, null checks
Readability Medium Clear naming, reasonable complexity
Security High No obvious vulnerabilities

Fail Conditions:

  • Doesn't follow existing patterns
  • Contains obvious bugs
  • Missing critical error handling
  • Security vulnerability present
4. Integration (Required)

Question: Will this work with existing code?

Checklist:

  • Imports are correct and exist
  • Exports match expected interface
  • No circular dependencies
  • Types are compatible (if TypeScript)

Fail Conditions:

  • Broken imports
  • Type mismatches
  • Circular dependency introduced
5. Completeness of Output (Optional)

Question: Is the output complete and usable?

Checklist:

  • No placeholder comments (// TODO, // FIXME)
  • No incomplete implementations
  • No ... or ellipsis in code
  • File contents are complete

Fail Conditions:

  • Contains placeholder code
  • Implementation is skeletal

Review Decision Matrix

Completeness Criteria Quality Integration Decision
✓ ✓ ✓ ✓ APPROVED
✓ ✓ ✗ ✓ NEEDS_REVISION
✓ ✗ ✓ ✓ NEEDS_REVISION
✓ ✓ ✓ ✗ NEEDS_REVISION
✗ - - - NEEDS_REVISION
- ✗✗ - - ESCALATE (if max retries)

Writing Review Feedback

For APPROVED
## Review: Task {id} - APPROVED ✓

All criteria met. Output is ready for verification.

**Summary**:
- Completeness: PASS
- Acceptance: {n}/{n} criteria met
- Quality: Good
- Integration: Compatible
For NEEDS_REVISION
## Review: Task {id} - NEEDS_REVISION ↻

Attempt {n} of {max}. Issues found:

**Issue 1**: {Specific issue}
- File: {path}
- Line: {number} (if applicable)
- Current: {what's wrong}
- Expected: {what's needed}
- Fix: {specific instruction}

**Issue 2**: {Specific issue}
- Fix: {specific instruction}

**Focus Areas for Retry**:
1. {Priority 1}
2. {Priority 2}
For ESCALATE
## Review: Task {id} - ESCALATE ⚠️

Max retries ({max}) reached or blocking issue found.

**Reason for Escalation**: {why}

**Attempts Summary**:
| Attempt | Issue | Feedback Given |
|---------|-------|----------------|
| 1 | {issue} | {feedback} |
| 2 | {issue} | {feedback} |
| 3 | {issue} | {feedback} |

**Options for User**:
1. Provide additional guidance and retry
2. Manually implement this task
3. Skip this task (if [OPTIONAL])
4. Modify requirements

Quality Standards

Code Pattern Matching

When reviewing, compare against:

  1. Files in the same directory
  2. Similar components/functions in codebase
  3. Patterns specified in design.md
  4. Project style guide (if exists)
Common Issues to Watch For

JavaScript/TypeScript:

  • Missing async/await
  • Unhandled promise rejections
  • Incorrect this binding
  • Missing type annotations

React:

  • Missing key props in lists
  • Direct state mutation
  • Missing dependency arrays in hooks
  • Memory leaks (missing cleanup)

API/Backend:

  • Missing input validation
  • SQL injection risks
  • Missing authentication checks
  • Improper error responses

Iteration Limits

Task Type Max Iterations Escalate After
Standard 3 3 failures
[CRITICAL] 5 5 failures
[OPTIONAL] 2 Skip after 2

使用示例

查看来源仓库中的使用示例与配置说明。

查看来源 ↗

来源历史

收录时间
2026-01-20
索引更新时间
2026-01-26
Branch
main

收录与同步日期属于本站记录。源码修改历史请查看来源仓库。

GitHub ↗

chat Comments (0)

chat_bubble_outline

No comments yet. Be the first to share your thoughts!