claude-tooling

Dynamically discover all Claude Code tooling available in this project and present it as a categorized summary.

Discovery steps

1. Project skills

Read all SKILL.md files under .claude/skills/. For each, extract the name and description from the YAML frontmatter. Note whether disable-model-invocation: true (user-only) or user-invokable: false (Claude-only) or neither (both).

2. Project agents

List all .md files under .claude/agents/. For each, read the first paragraph to get a one-line description.

3. Installed plugins

Run:

claude plugins list

List each installed plugin and its description.

4. MCP servers

Run:

claude mcp list

List each configured MCP server and how it was added (project vs user scope).

5. Hooks

Read .claude/settings.json and extract the hooks section. Summarize each hook: event type, matcher (if any), and what the command does in plain English.

Output format

Present the results grouped by category. Use this structure:

## Project Skills
| Skill | Description | Invocation |
|-------|-------------|------------|
| /skill-name | description | User / Claude / Both |

## Project Agents
| Agent | Description |
|-------|-------------|
| agent-name | one-line description |

## Installed Plugins
| Plugin | Description |
|--------|-------------|
| plugin-name | description |

## MCP Servers
| Server | Scope |
|--------|-------|
| server-name | project / user |

## Active Hooks
| Event | Trigger | Action |
|-------|---------|--------|
| PreToolUse | Edit, Write | Blocks edits to generated/sensitive files |

If a category has no entries, omit it entirely.