LiuYihey Task Review Guide
This skill guides you through reviewing completed tasks and making intelligent decisions about skill creation and updates.
When to Use This Skill
- After completing any task (automatically triggered by review_task tool)
- When evaluating whether a solution is worth capturing as a skill
- When assessing if a used skill performed well
Review Workflow
Step 1: Identify Skill Usage
Check if any skills were used during the task:
If skills were used: Proceed to Step 2 (Evaluate Skill Performance) If no skills were used: Proceed to Step 3 (Evaluate Reuse Potential)
Step 2: Evaluate Skill Performance (Skills Were Used)
Analyze how well the skill(s) performed:
| Performance | Indicators | Action |
|---|---|---|
| Excellent | Task completed smoothly, no workarounds needed, instructions were clear | No action needed. Acknowledge success. |
| Good with minor issues | Mostly worked but had small gaps or unclear parts | Consider updating the skill with update_skill tool |
| Poor | Required significant workarounds, instructions were wrong/incomplete | Strongly recommend updating the skill with update_skill tool |
Decision Criteria for Skill Updates
Update the skill if ANY of these apply:
- Instructions were unclear or incomplete
- Edge cases weren't handled
- Commands/APIs were outdated
- Had to improvise or work around issues
- Better patterns were discovered during execution
If update is needed: Call the update_skill MCP tool, which will provide the skill-updater guide.
Step 3: Evaluate Reuse Potential (No Skills Were Used)
Assess whether the completed task is worth capturing as a new skill:
High Reuse Value ✅ (Create Skill)
Create a skill if the task has:
- Repeatable workflow: Same steps apply to similar problems
- Clear trigger conditions: Easy to identify when this solution applies
- Non-trivial complexity: More than 3-4 steps or requires specific knowledge
- Broad applicability: Useful across multiple projects/contexts
Examples of high reuse value:
- Setting up a development environment (React + TypeScript + Tailwind)
- Implementing authentication flow (JWT + refresh tokens)
- Configuring CI/CD pipeline (GitHub Actions + Docker)
- Data processing workflow (CSV → clean → analyze → visualize)
Low Reuse Value ❌ (Don't Create Skill)
Don't create a skill if the task is:
- One-off solution: Highly specific to current context
- Trivial: Simple enough that agent can handle without guidance
- Too variable: Every instance requires completely different approach
- Already covered: Existing skills or general knowledge suffice
Examples of low reuse value:
- Fixing a specific bug in specific codebase
- Writing a one-time data migration script
- Answering a factual question
- Simple file operations (copy, rename, delete)
Decision Framework
Ask yourself:
- Would I do this exact workflow again? (If no → don't create)
- Is this more complex than basic agent knowledge? (If no → don't create)
- Can I clearly define when to use this? (If no → don't create)
- Would this save significant time in future tasks? (If no → don't create)
If 3+ answers are "yes": Create the skill with create_skill tool
If 2 or fewer "yes": Skip skill creation, acknowledge task completion
If skill creation is needed: Call the create_skill MCP tool, which will provide the skill-creator guide.
Review Response Template
Keep your review response concise. Use this template:
If Skills Were Used:
Task completed using skill(s): [skill-name(s)]
Performance: [Excellent/Good/Poor]
[If issues found]: Recommend updating skill due to [specific issue]
[Action]: [Call update_skill tool / No action needed]
If No Skills Were Used:
Task completed without skills.
Reuse potential: [High/Low]
Reasoning: [1-2 sentence justification]
[Action]: [Call create_skill tool / No skill creation needed]
Examples
Example 1: Skill Used Successfully
Task completed using skill: react-ts-setup
Performance: Excellent - all steps executed smoothly
No action needed.
Example 2: Skill Used with Issues
Task completed using skill: api-integration
Performance: Poor - missing error handling for 429 rate limits
Recommend updating skill to add retry logic with exponential backoff
Action: Calling update_skill tool
Example 3: No Skill, High Reuse Value
Task completed without skills.
Reuse potential: High
Reasoning: Implemented OAuth2 PKCE flow - repeatable pattern for secure authentication, 8+ steps, applicable to many projects
Action: Calling create_skill tool
Example 4: No Skill, Low Reuse Value
Task completed without skills.
Reuse potential: Low
Reasoning: Fixed specific CSS bug in current project - too context-specific, unlikely to recur
No skill creation needed.
Quality Checklist
Before making a decision, verify:
- Considered actual task complexity and repeatability
- Evaluated skill performance objectively (if skills were used)
- Identified specific issues (if recommending updates)
- Justified decision with concrete reasoning
- Kept response concise and actionable