seth-schultz/orchestr8/prompts

Create Skill: {{task}}

Request: {{task}}

Your Role

You are creating a new skill fragment documenting a specific technique or capability.

Phase 1: Skill Definition (0-25%)

→ Load: orchestr8://skills/match?query=skill+authoring+fragment&mode=index&maxResults=5

Activities:

• Define skill scope for {{task}}
• Identify when this skill applies (useWhen scenarios)
• Determine skill category (testing, deployment, architecture, etc.)
• Check for existing similar skills (avoid duplication)
• Plan granularity (focused on single capability)
• Define target token budget (400-600 tokens)

→ Checkpoint: Skill defined, no duplicates

Phase 2: Structure Planning (25-50%)

→ Load: orchestr8://skills/match?query=fragment+metadata&mode=index&maxResults=3

Activities:

• Plan content structure (technique, steps, examples)
• Design frontmatter metadata
• Select 4-6 relevant tags (lowercase, specific)
• Define 3-4 capabilities (what skill enables)
• Write 3-4 useWhen scenarios (when to apply)
• Plan practical examples (1-2 code examples)

→ Checkpoint: Structure planned

Phase 3: Content Creation (50-80%)

→ Load: orchestr8://match?query={{task}}&categories=skill,pattern,example&mode=index&maxResults=8

Activities:

• Write skill frontmatter (YAML metadata)
• Document the technique clearly
• Provide step-by-step instructions
• Include practical code examples
• Document prerequisites and dependencies
• Add common pitfalls and troubleshooting
• Include related skills references
• Ensure token estimate accurate

→ Checkpoint: Skill content written

Phase 4: Validation & Integration (80-100%)

→ Load: orchestr8://skills/match?query=testing+discovery&mode=index&maxResults=3

Activities:

• Test fuzzy matching with queries
• Verify metadata enables discovery
• Optimize tags if needed
• Save to /resources/skills/
• Test loading via orchestr8:// URI
• Verify token count matches estimate
• Rebuild resource index
• Test integration in workflows

→ Checkpoint: Skill discoverable, integrated

Success Criteria

✅ Focused on single technique or capability ✅ Appropriate size (400-600 tokens) ✅ Rich metadata (4-6 tags, 3-4 capabilities, 3-4 useWhen) ✅ Clear step-by-step instructions ✅ Practical examples included ✅ Discoverable via fuzzy matching ✅ Saved to correct location ✅ Index rebuilt successfully