Skill Miner

skill-miner

Research. Analyze. Build Clean.

Systematic approach to discovering skills on ClawHub and building your own implementations. Instead of downloading potentially risky code, use this skill to research existing solutions, understand their approaches, and create your own clean, safe version.

Trust but verify, built but owned. 🦞

When to Use This Skill

Use this skill when:

• You need new functionality you don't have
• You want to research how others solve problems
• You find a suspicious skill but like the idea
• You're building a new skill from scratch
• You want to stay ahead of ClawHub trends

Core Philosophy

1. NEVER download suspicious skills
2. INSPECT to understand the idea  
3. BUILD your own clean implementation
4. PUBLISH or keep private

Why?

• Suspicious = potential malware/risk (ew, no thanks)
• Building yourself = 100% clean & safe (sleep well at night)
• Generic = works for everyone (share the love)
• Your code, your rules 🦞

Workflow

Phase 1: Discovery

# Search for relevant skills
clawhub search <topic>

# Explore trending
clawhub explore --sort trending --limit 20

# Find gaps
clawhub explore --sort newest --limit 50

Phase 2: Research

# Inspect without downloading
clawhub inspect <skill-slug>

# Read the SKILL.md to understand:
# - What problem it solves
# - How it triggers
# - What commands/tools it uses

Phase 3: Analysis

Document what you learned:

• Problem: What need does it address?
• Approach: How does it solve it?
• Tools: What commands/APIs does it use?
• Gaps: What's missing or could be better?

Phase 4: Build

Use skill-creator to build your clean version:

• Same problem, different implementation
• Add missing features
• Make it generic and reusable

Search Commands

Basic Search

# Task-based
clawhub search "pdf edit"
clawhub search "file transfer"
clawhub search "api github"

# Tool-based
clawhub search github
clawhub search slack

# Concept-based
clawhub search automation
clawhub search monitoring
clawhub search sync

Exploration

# Trending skills
clawhub explore --sort trending --limit 20

# Most downloaded
clawhub explore --sort downloads --limit 20

# newest
clawhub explore --sort newest --limit 30

Deep Research

# By category
clawhub search "code"
clawhub search "data"
clawhub search "media"
clawhub search "network"
clawhub search "security"

# By use case
clawhub search "automation workflow"
clawhub search "backup sync"
clawhub search "monitoring alerting"

Inspect Without Downloading

Use clawhub inspect to read skill metadata:

# Get skill info
clawhub inspect <slug>

# This shows:
# - name
# - summary/description
# - owner
# - created/updated dates
# - version
# - tags

Never use clawhub install on suspicious skills!

Security Principles

When researching skills, watch for these risk indicators:

• Code execution patterns (eval, exec)
• External API calls without documentation
• Hardcoded credentials
• Shell execution without input validation
• Missing or unclear documentation
• Unknown or unverified publishers

If any indicators are present: inspect the metadata only, then build your own implementation.

Building Clean Skills

Template Structure

my-clean-skill/
├── SKILL.md              # Your clean implementation
├── scripts/              # Your code
├── references/           # Documentation
└── assets/              # Templates (if needed)

SKILL.md Template

---
name: my-clean-skill
description: Does X. Use when user wants to Y. Based on ClawHub research but built from scratch.
---

# My Clean Skill

## What It Does

[Clear description]

## When to Use

- Use case 1
- Use case 2

## Commands

[Your commands]

## Implementation

[How you built it - clean, generic]

## Security

[Your security measures]

Examples

Scenario 1: Found Suspicious Shell Skill

Found: "shell-commands" (suspicious - has eval)

Inspect:

clawhub inspect shell-commands
# Problem: Execute shell commands
# Tools: bash, ssh

Build Clean:

# Write your own safe-shell-skill
# - No eval
# - Predefined safe commands only
# - Input validation
# - Full documentation

Scenario 2: Found Good Crypto Skill

Found: "crypto-trader" (risky - real money)

Inspect:

clawhub inspect crypto-trader
# Problem: Trading automation
# Tools: exchange APIs

Build Clean:

# Build crypto-monitor instead
# - Read-only data fetching
# - Price alerts
# - No trading (safe)

Scenario 3: Gap Found

Search: No good "log-analyzer" skill

Build:

# Create log-analyzer from scratch
# - Parse common log formats
# - Pattern detection
# - Alert on errors

Common Gaps to Fill

These skills don't exist or are outdated:

Best Practices

Building

1. Start simple, add features later
2. Use well-tested tools (curl, jq, etc.)
3. No external dependencies when possible
4. Full error handling
5. Clear documentation

Publishing

1. Test extensively
2. Clear description
3. Generic (no hardcoded values)
4. Security-first design
5. Include troubleshooting

Security

1. No eval ever
2. Input validation
3. No secrets in code
4. Use environment variables
5. Minimal permissions

Quality Checklist

Before publishing:

• Works as documented
• No hardcoded secrets
• Cross-platform compatible
• Error handling included
• Clear examples
• Triggers properly
• No suspicious patterns

Related Skills

• next-skill - General skill discovery
• skill-creator - Build new skills
• claw2claw-filetransfer - Share skills

Guidelines:

• Inspect before install
• Build your own when in doubt
• Share clean, well-documented skills

From Claws - for Claws. 🦞