oyi77

find-skills

Automatically discover, evaluate, and activate community skills when local skills don't cover user needs. Includes credibility scoring and safety checks for complete OpenClaw self-sufficiency.

oyi77 0 Updated 3mo ago
GitHub

Install

npx skillscat add oyi77/1ai-skills/find-skills

Install via the SkillsCat registry.

SKILL.md

Find Skills - Intelligent Skill Discovery System

Overview

Automatically discover and integrate community skills from skills.sh and awesome-openclaw-skills when your local skills don't cover a need. Includes credibility scoring, safety checks, and seamless activation.

Makes your OpenClaw complete and self-sufficient - never say "I can't do that" again!

When to Use

Automatic Activation when:

  • User asks "how do I do X" where X isn't covered by local skills
  • User says "find a skill for X" or "is there a skill for X"
  • User asks "can you do X" where X is specialized
  • User wants to extend agent capabilities
  • User mentions wishing for help with a specific domain

Examples:

  • "How do I optimize my React app?" → Search for React performance skills
  • "Can you help with Docker?" → Search for Docker skills
  • "I need to create a changelog" → Search for changelog skills

Skill Discovery Process

Step 1: Check Local Skills First

function checkLocalSkills(userRequest) {
  // Check .skill-activation.json
  const localSkills = loadActivationRules();
  
  // Search by keywords
  const matches = findMatchingSkills(userRequest, localSkills);
  
  if (matches.length > 0) {
    return { found: true, skills: matches, source: 'local' };
  }
  
  return { found: false };
}

Step 2: Search Community Skills

# Search skills.sh
npx skills find [query]

# Example searches
npx skills find react performance
npx skills find docker deployment
npx skills find api testing

Search Strategy:

  1. Extract key terms from user request
  2. Search skills.sh first (curated, high quality)
  3. Fall back to awesome-openclaw-skills (comprehensive)
  4. Return top 5 results ranked by credibility

Step 3: Evaluate Credibility

Credibility Score (0-100):

function calculateCredibilityScore(skill) {
  let score = 0;
  
  // 1. Usage Stats (40 points)
  const usageCount = skill.usageCount || 0;
  if (usageCount > 100000) score += 40;
  else if (usageCount > 50000) score += 35;
  else if (usageCount > 10000) score += 30;
  else if (usageCount > 5000) score += 25;
  else if (usageCount > 1000) score += 20;
  else score += Math.min(usageCount / 50, 15);
  
  // 2. Author Reputation (30 points)
  const trustedAuthors = [
    'vercel-labs',
    'anthropics',
    'obra',
    'supabase',
    'remotion-dev',
    'expo',
    'better-auth'
  ];
  
  if (trustedAuthors.includes(skill.author)) {
    score += 30;
  } else if (skill.authorStars > 1000) {
    score += 25;
  } else if (skill.authorStars > 500) {
    score += 20;
  } else {
    score += Math.min(skill.authorStars / 25, 15);
  }
  
  // 3. Update Frequency (15 points)
  const daysSinceUpdate = getDaysSince(skill.lastUpdated);
  if (daysSinceUpdate < 30) score += 15;
  else if (daysSinceUpdate < 90) score += 12;
  else if (daysSinceUpdate < 180) score += 8;
  else if (daysSinceUpdate < 365) score += 5;
  else score += 2;
  
  // 4. Documentation Quality (15 points)
  if (skill.hasExamples) score += 5;
  if (skill.hasCodeSamples) score += 5;
  if (skill.descriptionLength > 500) score += 5;
  
  return Math.min(score, 100);
}

Safety Thresholds:

  • 90-100: Auto-install (trusted sources)
  • 70-89: Recommend with high confidence
  • 50-69: Recommend with caution
  • Below 50: Show but warn user

Step 4: Present to User

## Found Skills for [Query]

### 🌟 Highly Recommended (Score: 95/100)
**vercel-react-best-practices** by vercel-labs
- React and Next.js performance optimization
- 137.4K uses | Updated 2 days ago
- [View Details](https://skills.sh/vercel-labs/agent-skills/vercel-react-best-practices)

**Install**: `npx skills add vercel-labs/agent-skills@vercel-react-best-practices`

### ✅ Recommended (Score: 78/100)
**react-performance-tips** by community-dev
- React optimization patterns
- 8.2K uses | Updated 1 month ago
- [View Details](https://skills.sh/community-dev/skills/react-performance-tips)

### ⚠️ Use with Caution (Score: 55/100)
**react-optimizer** by new-dev
- Basic React optimization
- 450 uses | Updated 6 months ago
- [View Details](https://skills.sh/new-dev/skills/react-optimizer)

**Would you like me to install the top recommendation?**

Step 5: Auto-Install (with Permission)

# For high-credibility skills (90+)
npx skills add vercel-labs/agent-skills@vercel-react-best-practices -g -y

# -g: Install globally (user-level)
# -y: Skip confirmation prompts

After Installation:

  1. Update .skill-activation.json with new skill reference
  2. Reload skill activation rules
  3. Confirm to user that skill is now available

Integration with Activation Rules

Update .skill-activation.json

{
  "skillActivation": {
    "strategy": "selective",
    "coreSkills": [
      "core/agent-docs",
      "core/joko-orchestrator",
      "core/find-skills"
    ],
    "externalSkills": {
      "vercel-labs/agent-skills": {
        "skills": [
          "vercel-react-best-practices",
          "next-best-practices"
        ],
        "credibilityScore": 95,
        "installedDate": "2026-02-17",
        "autoUpdate": true
      },
      "obra/superpowers": {
        "skills": [
          "systematic-debugging",
          "test-driven-development"
        ],
        "credibilityScore": 92,
        "installedDate": "2026-02-17",
        "autoUpdate": true
      }
    },
    "fallbackBehavior": {
      "searchCommunity": true,
      "minCredibilityScore": 70,
      "autoInstallThreshold": 90,
      "askUserBelow": 90
    }
  }
}

Credibility Indicators

Trusted Sources (Auto-Install at 90+)

Author Reputation Why Trusted
vercel-labs ⭐⭐⭐⭐⭐ Official Vercel engineering team
anthropics ⭐⭐⭐⭐⭐ Official Anthropic (Claude) team
obra ⭐⭐⭐⭐⭐ Superpowers creator, trusted community leader
supabase ⭐⭐⭐⭐⭐ Official Supabase team
remotion-dev ⭐⭐⭐⭐ Official Remotion video team
expo ⭐⭐⭐⭐ Official Expo/React Native team

Red Flags (Warn User)

  • ⚠️ No updates in 1+ year
  • ⚠️ Less than 100 uses
  • ⚠️ Unknown author with no GitHub stars
  • ⚠️ No documentation or examples
  • ⚠️ Suspicious package name
  • ⚠️ Requests unusual permissions

Automation Examples

Example 1: User Asks for React Help

USER: "How do I make my React app faster?"

AGENT:
1. Check local skills → No React performance skill found
2. Search community → npx skills find react performance
3. Evaluate results:
   - vercel-react-best-practices (Score: 95)
   - react-performance-patterns (Score: 82)
4. Present top result to user
5. Offer to install

RESPONSE:
"I found a highly-rated skill for React performance optimization!

**vercel-react-best-practices** (Credibility: 95/100)
- Official Vercel engineering guidelines
- 137.4K uses, updated 2 days ago

Would you like me to install it? This will give me comprehensive React and Next.js optimization knowledge."

Example 2: User Needs Docker Help

USER: "Can you help me with Docker?"

AGENT:
1. Check local skills → No Docker skill found
2. Search → npx skills find docker
3. Find: docker-expert (Score: 92)
4. Auto-install (score > 90)
5. Update activation rules
6. Confirm to user

RESPONSE:
"✅ I've installed the 'docker-expert' skill (Credibility: 92/100)!

I can now help you with:
- Docker containerization
- Dockerfile optimization
- Docker Compose
- Container debugging
- Best practices

What Docker task would you like help with?"

Example 3: Low Credibility Skill

USER: "Find a skill for blockchain development"

AGENT:
1. Search → npx skills find blockchain
2. Find: crypto-dev-helper (Score: 48)
3. Below threshold (< 70)
4. Warn user

RESPONSE:
"⚠️ I found a skill for blockchain development, but it has low credibility:

**crypto-dev-helper** (Credibility: 48/100)
- Only 320 uses
- Last updated 8 months ago
- Unknown author

I recommend:
1. Searching for more established blockchain skills
2. Using general development skills instead
3. Proceeding with caution if you choose to install

Would you like me to search for alternatives?"

Safety Features

1. Sandboxed Evaluation

// Before installing, check skill metadata
function evaluateSkillSafety(skill) {
  const risks = [];
  
  // Check for suspicious patterns
  if (skill.requiresSystemAccess) {
    risks.push('Requires system-level access');
  }
  
  if (skill.hasExternalDependencies) {
    risks.push('Has external dependencies');
  }
  
  if (skill.modifiesGlobalState) {
    risks.push('Modifies global configuration');
  }
  
  return {
    safe: risks.length === 0,
    risks: risks
  };
}

2. User Confirmation

For scores 70-89:
"This skill looks good but I recommend reviewing it first.
[View on skills.sh] [Install Anyway] [Skip]"

For scores < 70:
"⚠️ This skill has low credibility. I don't recommend installing it.
[View Details] [Install at Your Own Risk] [Cancel]"

3. Rollback Capability

# If skill causes issues
npx skills remove vercel-labs/agent-skills@vercel-react-best-practices

# Restore previous activation rules
git checkout .skill-activation.json

Workflow Integration

Daily Routine Assistant

// Morning routine
async function morningRoutine() {
  // Check for skill updates
  await run('npx skills check');
  
  // Update high-credibility skills
  const updates = await getAvailableUpdates();
  for (const skill of updates) {
    if (skill.credibilityScore >= 90) {
      await run(`npx skills update ${skill.name}`);
    }
  }
  
  // Report to user
  console.log('✅ Skills updated and ready for the day!');
}

Continuous Learning

// Track what skills are used most
function trackSkillUsage() {
  // Log skill activations
  // Identify gaps
  // Suggest new skills proactively
  
  if (userFrequentlyAsksAbout('testing')) {
    suggestSkill('javascript-testing-patterns');
  }
}

Best Practices

1. Always Check Local First

Don't search community if you already have the capability.

2. Prefer High-Credibility Sources

Trust vercel-labs, anthropics, obra over unknown authors.

3. Keep Skills Updated

# Weekly update check
npx skills check
npx skills update

4. Monitor Skill Performance

  • Track which skills are actually helpful
  • Remove unused skills
  • Report issues to skill authors

5. Contribute Back

  • If you create useful skills, publish them
  • Share improvements with community
  • Help others discover your Indonesian-specific skills

Commands Reference

# Search for skills
npx skills find [query]

# Install skill
npx skills add <owner/repo@skill> -g -y

# Check for updates
npx skills check

# Update all skills
npx skills update

# Remove skill
npx skills remove <owner/repo@skill>

# List installed skills
npx skills list

Related Skills

  • core/agent-docs - Documentation for agent capabilities
  • core/joko-orchestrator - Orchestrate multiple skills
  • All your custom skills (payment, social media, etc.)

Last Updated: 2026-02-17
Purpose: Make OpenClaw complete and self-sufficient
Key Feature: Intelligent skill discovery with credibility scoring