Skill Discovery

Scan and catalog all available skills, agents, commands, hooks, and rules in the project. Produces a structured skills-registry.json and presents a formatted summary.

Steps

Step 1: Run the Scanner

Execute the registry scanner script from the repo root:

cd "D:/ailocal/acm-ai" && bash .claude/skills/skill-discovery/scripts/scan_registry.sh

This generates skills-registry.json at the repo root with a full inventory of all capabilities.

Step 2: Read the Generated Registry

Read the generated file:

D:/ailocal/acm-ai/skills-registry.json

Step 3: Present a Formatted Summary

Present the results grouped by type. Use this format:

## Skills Registry Summary
Scanned at: <scanned_at>

### Skills (<count> total)
| Name | Description | Location | Platforms |
|------|-------------|----------|-----------|
| pydantic-models-py | Create Pydantic models... | .claude/skills/... | claude-code |
| ...  | ...          | ...      | ...       |

### Commands (<count> total)
| Name | Description |
|------|-------------|
| start | Start development services |
| ...  | ...         |

### Hooks (<count> total)
| Name | Trigger |
|------|---------|
| pre-tool-use.sh | PreToolUse |
| ...             | ...        |

### Rule Categories (from CLAUDE.md)
- Project Overview
- Essential Commands
- Architecture
- ...

Step 4: Filter if Requested

If the user asks to filter (e.g., "show me debugging skills", "list only frontend skills"):

• Filter skills array by matching keywords against name and description fields
• Present the filtered subset using the same table format
• Note how many skills were hidden: "Showing 3 of 12 skills (filtered by 'debug')"

Quick Reference

The following skills are available in this project (populated after first scan):

Run the scanner to get the full up-to-date list including all installed skills.

Notes

• Skills can live in .claude/skills/ (Claude Code) or .agents/skills/ (agent platform) or both
• Skills with also_at entries appear in multiple registries — they are the same skill
• The registry is regenerated on each scan; commit skills-registry.json to track changes over time
• Hooks are found in .claude/hooks/ — they run automatically based on trigger type
• Commands are in commands/ (surreal-commands worker handlers, not slash commands)