菜单
Optimization patterns and grading system for skill quality assessment. Automatically loaded when creating new skills, evaluating skill quality, reviewing skill content, or when "skill optimization", "skill quality", "BP pattern", "skill grading", or "skill assessment" are mentioned.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
This skill provides templates, analysis frameworks, and quality criteria for game reverse-engineering and competitive analysis. Used when analyzing external games for mechanics, technology, marketing, retention, and market positioning. Loaded by game-researcher agent and game-analyze command.
This skill guides subagent coordination through game development workflows. Automatically loaded when orchestrating multiple agents, managing workflow phases, determining autonomous execution mode, or when "orchestration", "workflow phases", "scale determination", "stop points", or "autonomous mode" are mentioned.
Advanced Jobs To Be Done methodology knowledge base. Use when analyzing products, markets, customer segments, or making product decisions. Provides access to complete AJTBD framework and expert prompts.
Product execution frameworks for converting strategy analysis into actionable product plans. Use when creating feature specifications, product roadmaps, MVP definitions, opportunity maps, or when "feature spec", "roadmap", "MVP", "product plan", "opportunity map", "WSJF", "Kano", "Shape Up", or "story mapping" is mentioned.
This skill should be used when creating strategy documents — context briefs, market analyses, competitive landscapes, strategy canvases, business models, GTM plans, pricing analyses, growth plans, initiative prioritizations, or strategic reports. Provides document templates, creation decision matrix, and quality standards.
Orchestration rules for strategy analysis workflows. Loaded when coordinating strategy agents, managing analysis phases, or running /strategy-report. Defines flows, stop points, agent coordination, and report compilation.
| name | skill-optimization |
| description | Optimization patterns and grading system for skill quality assessment. Automatically loaded when creating new skills, evaluating skill quality, reviewing skill content, or when "skill optimization", "skill quality", "BP pattern", "skill grading", or "skill assessment" are mentioned. |
BP-1: Negative-to-Positive Rewriting
Convert prohibitions into actionable instructions. Agents follow positive directives more reliably than avoiding negatives.
BAD: "Don't use any in TypeScript"
GOOD: "Use explicit types: unknown for external data, generics for reusable code, union types for variants"
BP-2: Vague-to-Specific Conversion
Replace ambiguous guidance with measurable criteria.
BAD: "Write clean functions"
GOOD: "Function length: <50 lines ideal, <200 max. Single return type. Max 3 parameters (use options object beyond 3)."
BP-3: Missing Output Format Specification
Every skill used by an agent that produces structured output MUST define the expected format.
BAD: "Report the findings"
GOOD: "Output JSON: { status, findings: [{category, severity, location, remediation}], summary }"
BP-4: Unstructured-to-Organized
Content should flow from general principles to specific patterns. Use consistent heading hierarchy.
BAD: Mixed tips, rules, and examples in random order
GOOD: Principle → When to apply → Pattern → Example → Anti-pattern
BP-5: Missing Context Provision
State when and why a rule applies, not just what the rule is.
BAD: "Use Result type for error handling"
GOOD: "Use Result type when: crossing service boundaries, parsing external input, operations with expected failure modes. Skip for: internal logic with guaranteed preconditions."
BP-6: Complex-to-Decomposed
Break complex multi-part rules into discrete, independently applicable items.
BAD: "Handle errors properly with logging, user notification, and recovery"
GOOD: Three separate sections: "Error Logging" (what/when/format), "User Notification" (messages/severity), "Error Recovery" (retry/fallback/escalation)
BP-7: Biased Examples
Provide examples that cover both success and failure paths, not just happy paths.
BAD: Only showing successful validation
GOOD: Show validation success AND common validation failure scenarios
BP-8: Missing Escalation Criteria
Define when the agent should stop and escalate to the user instead of continuing.
BAD: No mention of edge cases or uncertainty handling
GOOD: "Escalate when: security implications unclear, performance impact >10%, breaking change to public API"
For detailed editing principles (context efficiency, deduplication, grouping, measurability, positive form, consistent notation, explicit prerequisites, priority ordering, scope boundaries), reference rule-editing-guide skill.
Context Efficiency: Skill SKILL.md should be self-contained within its scope. Reference other skills by name, never duplicate content.
Deduplication: If content overlaps with another skill, reference it: "See [skill-name] skill '[section]' for details."
Measurability: Every criterion should be checkable — "Is this rule followed? Yes/No" must be answerable from the code.
Scope Boundaries: First paragraph of skill MUST state what is in scope and what is not.
Expert References: Every skill should end with "Expert References (Reasoning Calibration)" table for decision grounding.
When evaluating a skill, check each item: