check-my-skills

check-my-skills — Skill Instruction Inspector

Purpose

Parse every skill in a project into a directed acyclic graph (DAG) of instruction nodes, visualize the flow as an interactive Dagre tree, and surface quality issues in the instruction chain. Two modes:

• Advanced (default): LLM-powered parsing via Anthropic API — deeper, catches nuance
• Standard (--easy): Heuristic parsing — fast, no API calls, works offline

---

Step 0 — Setup

Detect Mode

Check the user's request for --easy or easy flag.

• If present → set MODE=standard
• If absent → set MODE=advance

Locate Skill Directory

Find where this skill's supporting files are installed:

SKILL_DIR=$(find "$(pwd)" ~/.agents/skills -path '*/check-my-skills/SKILL.md' -type f 2>/dev/null -exec dirname {} \; | head -1)

Ensure Dependencies

python3 -c "import yaml" 2>/dev/null || pip3 install --user "PyYAML>=6.0"
python3 -c "import anthropic" 2>/dev/null || pip3 install --user "anthropic>=0.25.0"

If the Anthropic SDK install fails or ANTHROPIC_API_KEY is not set, fall back to standard mode automatically and inform the user. Or the user can explicitly use --easy to skip the API entirely.

---

Step 1 — Discovery

Scan for all skill directories. Search these locations in order:

# Project skills (current working directory)
find "$(pwd)" -name "SKILL.md" -path "*/skills/*" -type f 2>/dev/null

# User-level skills
find ~/.agents/skills -name "SKILL.md" -type f 2>/dev/null

# Also check if user specified a specific path
# e.g., "check my skills in /path/to/project"

For each SKILL.md found, read the file and extract:

• name (from YAML frontmatter)
• description (from YAML frontmatter)
• path (absolute path)
• compatibility / mcp_servers (from frontmatter if present)
• List of referenced files (e.g., references/examples.md, other skills mentioned)

Present the inventory to the user:

Found N skills:
1. product-brief (/mnt/skills/user/product-brief/SKILL.md)
2. prototype-prd (/mnt/skills/user/prototype-prd/SKILL.md)
...