编程
The ultimate meta-skill for OpenClaw. Automatically analyzes, diagnoses, and upgrades ANY existing skill to elite/legendary tier. Use this skill when you need to enhance, refactor, or evolve any skill — whether it's writing, coding, research, analysis, or custom skills. This skill injects upgrade protocols, quality gates, and enhancement layers that transform basic skills into world-class cognitive systems. Think of it as a "skill compiler" that optimizes every aspect of another skill's architecture, reasoning depth, output quality, and execution precision.
The ultimate meta-skill for OpenClaw. Automatically analyzes, diagnoses, and upgrades ANY existing skill to elite/legendary tier. Use this skill when you need to enhance, refactor, or evolve any skill — whether it's writing, coding, research, analysis, or custom skills. This skill injects upgrade protocols, quality gates, and enhancement layers that transform basic skills into world-class cognitive systems. Think of it as a "skill compiler" that optimizes every aspect of another skill's architecture, reasoning depth, output quality, and execution precision.
Mission: Take ANY skill from "functional" to "legendary". No skill is complete until it passes the Upgrader.
When asked to upgrade, enhance, or improve ANY skill, execute the SKILL UPGRADE PROTOCOL:
PHASE 1: DIAGNOSIS
Read the target skill completely
Identify: strengths, gaps, weaknesses, missed opportunities
Score across 10 dimensions (1-10)
PHASE 2: ARCHITECTURE UPGRADE
Restructure for clarity and power
Add missing components
Remove redundancy and fluff
PHASE 3: REASONING UPGRADE
Inject deep reasoning protocols
Add self-correction loops
Install verification frameworks
PHASE 4: OUTPUT UPGRADE
Elevate output standards
Add quality gates
Install polish protocols
PHASE 5: META-UPGRADE
Add inter-skill compatibility
Install context intelligence
Add error recovery
PHASE 6: FINAL VALIDATION
Run the skill through stress tests
Verify all upgrade claims
Produce the upgraded version
Skill ini membantu mengupgrade skill LAIN, tapi punya batas keras agar tidak merusak sistem:
openclaw.json, models.json berisi apiKey, env berisi token, _meta.json selain field version/slug).done lanjut ke berikutnya. Jangan rombak banyak sekaligus tanpa konfirmasi Bos.Before upgrading, score the target skill on ALL 10 dimensions:
D1. PURPOSE CLARITY (1-10)
Is the skill's purpose crystal clear?
Does it know EXACTLY when to activate?
Is the trigger condition unambiguous?
D2. REASONING DEPTH (1-10)
Does it think before acting?
Are there multi-step reasoning protocols?
Does it self-critique?
D3. FRAMEWORK RICHNESS (1-10)
How many proven frameworks does it include?
Are they from world-class sources?
Is there a selection guide?
D4. OUTPUT QUALITY (1-10)
Are output standards defined?
Is there a quality checklist?
Does it specify "good vs. great"?
D5. EDGE CASE HANDLING (1-10)
Does it handle errors gracefully?
Are there fallback protocols?
Does it anticipate failure modes?
D6. CONTEXT AWARENESS (1-10)
Does it manage context intelligently?
Can it handle large inputs?
Does it track state across turns?
D7. TOOL INTEGRATION (1-10)
Does it use tools optimally?
Are there tool selection rules?
Is error recovery covered?
D8. VOICE & PERSONA (1-10)
Is the voice consistent?
Is the persona well-defined?
Does it adapt tone to context?
D9. REFERENCE QUALITY (1-10)
Are references comprehensive?
Are they from authoritative sources?
Are they well-organized?
D10. COMPLETENESS (1-10)
Does it cover ALL aspects of the domain?
Are there templates?
Is there a quick-start guide?
Scoring Interpretation:
Apply these 7 layers to EVERY skill upgrade:
LAYER 1: CLARITY ENHANCEMENT
Restructure for instant comprehension
Add visual hierarchy (emojis, tables, diagrams)
Ensure every section has a "why this matters"
LAYER 2: DEPTH ENHANCEMENT
Add reasoning protocols (see super-intelligence-skill)
Install self-correction checkpoints
Add verification requirements
LAYER 3: BREADTH ENHANCEMENT
Expand coverage of the domain
Add edge cases and corner cases
Include "what if" scenarios
LAYER 4: PRECISION ENHANCEMENT
Remove ambiguity and vagueness
Add specific examples for every concept
Define success criteria explicitly
LAYER 5: ELEGANCE ENHANCEMENT
Refine voice and tone
Improve flow and rhythm
Add memorable phrases and mnemonics
LAYER 6: ROBUSTNESS ENHANCEMENT
Add error handling protocols
Install fallback strategies
Add "when this doesn't work" sections
LAYER 7: EXTENSIBILITY ENHANCEMENT
Design for future additions
Add extension points
Document how to customize
Every skill should evolve through these stages. Ensure the upgraded skill supports the full path:
STAGE 1: NOVICE
"Tell me what to do"
Needs step-by-step instructions
Requires examples for everything
STAGE 2: COMPETENT
"I know the basics"
Can follow frameworks independently
Makes predictable outputs
STAGE 3: PROFICIENT
"I can adapt"
Chooses frameworks based on context
Handles common variations
STAGE 4: EXPERT
"I can innovate"
Combines frameworks creatively
Anticipates edge cases
STAGE 5: MASTER
"I can teach"
Explains WHY, not just WHAT
Creates new frameworks
STAGE 6: LEGENDARY
"I transcend the framework"
Intuitively knows the best approach
Produces outputs that surprise and delight
The upgraded skill must support ALL stages. Novices get hand-holding. Masters get power tools.
Before declaring an upgrade complete, verify against these gates:
GATE 1: THE "SO WHAT?" GATE
Every section must answer: "So what? Why does this matter?"
If a section fails, rewrite or remove.
GATE 2: THE "CAN A NOVICE FOLLOW?" GATE
Give the skill to someone with zero context
Can they understand and execute?
If not, add prerequisites and scaffolding.
GATE 3: THE "CAN A MASTER USE?" GATE
Give the skill to an expert
Does it offer advanced techniques?
Does it respect their expertise?
GATE 4: THE "STRESS TEST" GATE
Throw edge cases at the skill
Does it handle gracefully?
Are failure modes documented?
GATE 5: THE "OUTPUT EXCELLENCE" GATE
Generate sample outputs using the skill
Are they truly elite-tier?
Would a human expert be impressed?
GATE 6: THE "INTEGRATION" GATE
Does the skill play well with other skills?
Are there clear handoff points?
Does it avoid conflicts?
During upgrade, ruthlessly eliminate these patterns:
VAGUE INSTRUCTIONS
"Write well" "Write with active voice, 8th-grade readability, and one idea per paragraph"
GENERIC ADVICE
"Be creative" "Generate 3 divergent approaches, evaluate each against [criteria]"
MISSING CONTEXT
"Analyze this" "Analyze using [framework], considering [constraints], output in [format]"
NO QUALITY BAR
"Create a report" "Create a report meeting [specific standards] with [verification steps]"
STATIC THINKING
"Follow these steps" "Follow these steps, but adapt based on [feedback loops]"
SINGLE-PATH REASONING
"Do X then Y" "Do X, verify result, if [condition] do Y, else do Z"
NO ERROR HANDLING
"Use this tool" "Use this tool, if it fails try [fallback], if still failing [escalation]"
FLUFF AND FILLER
Cut any sentence that doesn't add unique value
Cut any section that restates the obvious
Cut any example that doesn't illuminate
| File | Content |
|---|---|
{baseDir}/references/skill-anatomy.md | The anatomy of a legendary skill — what separates good from great |
{baseDir}/references/enhancement-layers.md | Deep dive into each of the 7 enhancement layers |
{baseDir}/references/skill-evolution.md | How skills evolve from basic to legendary |
{baseDir}/references/quality-standards.md | The absolute highest standards for skill output |
{baseDir}/templates/skill-upgrade-template.md | Ready-to-use template for upgrading any skill |
"Upgrade my writing skill to legendary tier"
"Enhance my coding skill with error recovery"
"Evolve my research skill to handle edge cases"
"Make my analysis skill produce expert-level outputs"
"Upgrade skill X to match the quality of skill Y"
Suggest reusable skills from recurring patterns in local memory files. Human review gate, drafts only to skills/_pending/, local-first runner with optional external fallback. Triggers on "skill forge", "propose a skill", "what skills should I have", "skill candidates", "what patterns have I been doing", "forge me a skill", "forge show", "forge accept", "forge reject", "forge promote".
Agent skill quality checker. Input a skill directory or skill files; output trigger clarity, metadata issues, examples, safety boundaries, installability, po...
Analyze the current conversation history and local installed skills to identify missed skill triggers, overlapping or duplicate skills, weak metadata, stale...
Create or repair local skills from verified, reusable workflows. Use after a non-obvious fix, a recurring procedure, a stale skill, or a request to save a workflow as a skill.
改任何文件前先看 diff、备份、等用户确认，安全升级已安装的代理技能。
Agent skill recommender. Input a user need, task description, or existing skill list; output best matching skills, install rationale, duplicate/merge candida...