SkillFather

Multi-Platform Agent Skill Fitness Analyzer

English · 简体中文

Don't install a skill just because it looks cool — find out if it's actually built for you.

---

Scope Statement: SkillFather focuses exclusively on fitness/suitability assessment — evaluating whether a Skill matches your actual use cases. It does not perform security audits, compliance checks, or vulnerability analysis.

What it does

Every Agent platform has a marketplace full of Skills. Some are brilliant — for someone else. SkillFather reads a Skill definition, asks you the right questions, and scores it across 5 dimensions so you know before installing whether it's worth your time.

SKILL.md → Parse → Generate 6-10 Diagnostic Questions → 5-Dimension Score → Fit Rating /10

The 5 Dimensions

Dimension	Weight	What it measures
Use-Case Fit	25%	Does the Skill's trigger scope cover your actual needs?
Environment Readiness	20%	Do you have the required tools, connectors, and dependencies?
Prerequisites	20%	Do you meet the Skill's preconditions and access requirements?
Workflow Match	20%	Does the Skill's execution flow match how you actually work?
Documentation Quality	15%	Is the Skill's documentation clear, complete, and actionable?

Supported Platforms

Platform	Skill Format	Storage Path
WorkBuddy	SKILL.md (YAML + Markdown)	~/.workbuddy/skills/
CodeBuddy	SKILL.md (YAML + Markdown)	~/.codebuddy/skills/
OpenAI Codex Agent	SKILL.md + agents/openai.yaml	.agents/skills/
Claude Code	.claude/commands/*.md	.claude/commands/
Coze	.skill / .zip / JSON	Web UI + exported files
Hermes Agent	SKILL.md (Hermes metadata)	~/.hermes/skills/

Quick Start

# Clone and install (zero external dependencies)
git clone https://github.com/alpsmonkey/SkillFather.git
cd SkillFather
pip install -e .

# Or with dev dependencies (pytest, ruff) for contributing
pip install -e ".[dev]"

Basic Usage

安装为 Agent 技能后，直接用自然语言触发即可，无需手动运行命令：

帮我分析一下这个技能
帮我分析一下这个skill
分析这个技能
分析这个skill
分析技能 / 评审技能 / 适配度分析

也可以通过命令行直接调用：

# List all supported platforms
skillfather platforms

# Auto-detect platform and analyze
skillfather analyze path/to/SKILL.md

# Specify a platform explicitly
skillfather analyze path/to/command.md --platform claude-code

# Interactive Q&A mode (get a personalized score)
skillfather analyze path/to/SKILL.md --interactive

# Output HTML report
skillfather analyze path/to/SKILL.md --format html -o report.html

# Output Markdown report
skillfather analyze path/to/SKILL.md --format markdown -o report.md

Platform-Specific Examples

# WorkBuddy / CodeBuddy
skillfather analyze ~/.workbuddy/skills/my-skill/SKILL.md
skillfather analyze ~/.codebuddy/skills/pdf-editor/SKILL.md --platform codebuddy

# OpenAI Codex Agent
skillfather analyze .agents/skills/deploy/SKILL.md --platform codex

# Claude Code
skillfather analyze .claude/commands/review.md --platform claude-code

# Coze / 扣子
skillfather analyze exported_skill.json --platform coze

# Hermes Agent
skillfather analyze ~/.hermes/skills/research/SKILL.md --platform hermes

Dual-Engine Architecture

Mode	Description	Dependencies
Rule Engine	Template-based offline analysis, zero config	None (pure stdlib)
LLM Enhanced	Uses an LLM to generate personalized questions and scoring	Any OpenAI-compatible API

LLM Mode (Optional)

# Set API key via environment variable
export SKILLFATHER_API_KEY=sk-xxx
skillfather analyze SKILL.md --llm --interactive

# Or use a custom API endpoint
export SKILLFATHER_API_BASE=https://api.deepseek.com/v1
export SKILLFATHER_MODEL=deepseek-chat
skillfather analyze SKILL.md --llm --interactive

Output Formats

Terminal

╔══════════════════════════════════════════════════╗
║  SkillFather — Fitness Analysis Report          ║
╠══════════════════════════════════════════════════╣
║  Skill:   SAP Cost Analysis                      ║
║  Platform: workbuddy                             ║
║  Score:   7.2 / 10                              ║
║  Verdict: RECOMMENDED                            ║
╠══════════════════════════════════════════════════╣
║  Use-Case Fit       ████████░░  8.0/10          ║
║  Environment        ██████░░░░  6.0/10          ║
║  Prerequisites      ███████░░░  7.0/10          ║
║  Workflow Match     ████████░░  7.5/10          ║
║  Documentation      ███████░░░  7.0/10          ║
╚══════════════════════════════════════════════════╝

HTML Report

Generate a styled, responsive HTML report with visual score breakdowns:

skillfather analyze SKILL.md --format html -o report.html

Markdown Report

For note-taking or sharing:

skillfather analyze SKILL.md --format markdown -o report.md

Configuration

Environment Variables

Variable	Description	Default
SKILLFATHER_API_KEY	LLM API Key	None (rule mode only)
SKILLFATHER_API_BASE	API Base URL	https://api.openai.com/v1
SKILLFATHER_MODEL	LLM model name	gpt-4o-mini
SKILLFATHER_CONFIG	Config file path	None

JSON Config (Optional)

Create skillfather.json in your project root or specify via SKILLFATHER_CONFIG:

{
  "llm": {
    "enabled": true,
    "api_key": "sk-xxx",
    "api_base": "https://api.openai.com/v1",
    "model": "gpt-4o-mini"
  },
  "analysis": {
    "num_questions": 10,
    "score_weights": {
      "use_case": 0.25,
      "environment": 0.2,
      "prerequisites": 0.2,
      "workflow": 0.2,
      "documentation": 0.15
    }
  }
}

Project Structure

SkillFather/
├── src/skillfather/
│   ├── __init__.py
│   ├── __main__.py
│   ├── cli.py              # CLI entry point (multi-platform)
│   ├── parser.py           # Universal Skill parser
│   ├── analyzer.py         # Analysis engine (rule + LLM)
│   ├── reporter.py         # HTML / Markdown reports
│   ├── config.py           # Configuration management
│   └── platforms/          # Platform adapters
│       ├── __init__.py     # Platform registry & auto-detection
│       ├── base.py         # Adapter base class
│       ├── workbuddy.py    # WorkBuddy
│       ├── codebuddy.py    # CodeBuddy
│       ├── codex.py        # OpenAI Codex Agent
│       ├── claude_code.py  # Claude Code
│       ├── coze.py         # Coze
│       └── hermes.py       # Hermes Agent
├── templates/
│   └── report.html         # HTML report template
├── examples/              # Sample files per platform
│   ├── sample_skill.md
│   ├── sample_codex_skill.md
│   ├── sample_claude_command.md
│   ├── sample_coze_skill.json
│   └── sample_hermes_skill.md
├── tests/
│   └── test_parser.py
├── dist/                    # Platform-specific SkillFather skill files
│   ├── claude-code/
│   │   └── skillfather.md   # ~/.claude/commands/skillfather.md
│   ├── codebuddy/
│   │   └── SKILL.md         # ~/.codebuddy/skills/skillfather/
│   ├── codex/
│   │   ├── SKILL.md         # ~/.agents/skills/skillfather/
│   │   └── agents/
│   │       └── openai.yaml
│   └── hermes/
│       └── productivity/
│           └── SKILL.md     # ~/.hermes/skills/productivity/skillfather/
├── SKILL.md                 # WorkBuddy skill definition (also used as root)
├── docs/
│   └── README_CN.md        # Chinese README
├── README.md
├── LICENSE
├── pyproject.toml
└── requirements.txt

Use as a Skill on Your Platform

SkillFather can be installed as a native skill on most major Agent platforms:

# Claude Code - slash command
cp dist/claude-code/skillfather.md ~/.claude/commands/skillfather.md

# CodeBuddy
cp dist/codebuddy/SKILL.md ~/.codebuddy/skills/skillfather/SKILL.md

# Codex (OpenAI Agent)
cp -r dist/codex ~/.agents/skills/skillfather

# Hermes Agent
cp -r dist/hermes/productivity/skillfather ~/.hermes/skills/productivity/skillfather

# WorkBuddy
cp SKILL.md ~/.workbuddy/skills/skillfather/SKILL.md

Note: Coze is a web-based platform — create a Bot manually via the Coze UI.

Why "Father"?

Every Skill has a creator, but nobody tells you whether that Skill was built for you. SkillFather is the judge — it reads the fine print, asks the hard questions, and gives you a number you can trust.

Install smarter. Not more.

License

MIT — use it, modify it, ship it.

Star History