Skill Activation Patterns

Skill Activation Patterns

Automatic skill suggestions based on user prompt keywords.

Two Scopes

Plugin-level (Recommended for plugins with multiple skills)

Plugins with related skills SHOULD include skill-activation to help users discover relevant skills:

my-plugin/
├── hooks/hooks.json              ← UserPromptSubmit hook
├── scripts/skill-activation.py   ← Hook script
├── .claude/skills/skill-rules.json  ← Plugin's own rules
└── skills/
    ├── skill-a/
    └── skill-b/

Example: skillmaker uses plugin-level activation to suggest skill-design when user mentions "create skill".

Project-level (User-configured)

Users can add skill-activation to their own projects independently:

my-project/
├── .claude/
│   └── skills/
│       └── skill-rules.json   ← User's rules
└── settings.json              ← UserPromptSubmit hook

Quick Start

1. Create .claude/skills/skill-rules.json
2. Add UserPromptSubmit hook
3. Hook reads rules, matches triggers, suggests skills

Core Concept

User Prompt → [Hook] → skill-rules.json → Match triggers → Suggest skills

skill-rules.json (Minimal)

{
  "version": "1.0",
  "skills": {
    "backend-patterns": {
      "type": "domain",
      "enforcement": "suggest",
      "promptTriggers": {
        "keywords": ["backend", "API"]
      }
    }
  }
}

Skill Types

Enforcement Levels

Best Practices

1. Start with suggest - Don't block until proven
2. Specific keywords - Avoid generic over-triggering
3. Test regex - Verify no false positives
4. Use skipConditions - Allow escape hatch

References

• Full Schema - Complete skill-rules.json spec
• Hook Implementation - TypeScript/Bash code
• Real Examples - Production configs