Component Scaffolding Protocol

Core Principle

Transform natural language specifications into production-ready plugin components. Generate complete, standards-compliant file structures with proper YAML frontmatter, following the 2026 Universal Agentic Runtime architecture.

Execution Framework

1. Parse Component Specification

Extract from user input:

• Component Type: skill, agent, or command
• Component Name: lowercase-with-hyphens (max 64 chars)
• Language/Framework: python, typescript, javascript, etc.
• Complexity Level: minimal, standard, or complex
• Special Requirements: hooks, allowed-tools, specific templates

Default Values (when unspecified):

• Location: project-level (.claude/)
• Language: based on project context
• Complexity: standard
• Permission Mode: inherited from environment

2. Apply Universal Agentic Runtime Standards

Generate compliant components following:

For Skills:

• SKILL.md with proper frontmatter (name, description, allowed-tools)
• Description following Discovery Tiering Matrix
• context: fork for isolated execution
• agent binding when appropriate
• Directory structure: skill-name/SKILL.md (+ scripts/, references/, assets/ if needed)

For Agents:

• Markdown file with YAML frontmatter
• name: lowercase-with-hyphens
• description: role definition with "MUST USE when" trigger
• tools: explicit whitelist (never omit)
• skills: auto-load relevant skills
• model: omit unless specified (defaults to configured subagent model)

For Commands:

• Markdown file with YAML frontmatter
• description: clear purpose with natural language triggers
• $ARGUMENTS placeholder for user input
• argument-hint: UI documentation
• allowed-tools: safety restrictions

3. Template Selection Logic

Minimal Complexity:

• Single component, straightforward purpose
• Standard frontmatter only
• No additional directories

Standard Complexity:

• Single component with standard features
• May include references/ for documentation
• Standard templates from existing bootstrap skills

Complex Complexity:

• Multi-component system
• references/, assets/, scripts/ directories as needed
• Router pattern or progressive disclosure patterns
• Multiple templates combined

4. Frontmatter Generation

Generate valid YAML with:

• All required fields populated
• Appropriate optional fields based on component type
• No redundant defaults (e.g., user-invocable: true)
• Proper tool permissions and restrictions

5. File Creation

Create files in correct locations:

• Skills: .claude/skills/skill-name/SKILL.md (or plugin-level)
• Agents: .claude/agents/agent-name.md
• Commands: .claude/commands/command-name.md

Write complete, production-ready files with:

• Proper YAML frontmatter
• Complete content following standards
• No TODO markers or placeholder text
• Ready to use immediately

Quality Gates

Generated components must:

• Pass YAML validation
• Follow naming conventions
• Include proper descriptions for semantic discovery
• Have appropriate tool restrictions
• Be self-contained (no cross-component coupling)
• Match 2026 Universal Agentic Runtime standards

Output Format

Return structured summary:

• Component Type and Name
• Location created
• Files generated
• Key features enabled
• Next steps (if any)

Example Output:

✓ Created skill: database-validator
  Location: .claude/skills/database-validator/SKILL.md
  Features: context: fork, agent: plugin-expert, allowed-tools: [Read, Grep]
  Status: Ready to use

Error Handling

If component exists:

• Report "Component already exists at [location]"
• Offer to overwrite or create with different name
• Never proceed without explicit confirmation

If invalid specification:

• Use reasonable defaults
• Apply standards-based decisions
• Never block on missing optional details

If directory creation fails:

• Verify path permissions
• Use project-level as fallback
• Report specific failure reason

Standards References

Core Architecture:

• CLAUDE.md Part IV (SKILL Protocol)
• CLAUDE.md Part III (AGENT Architecture)
• CLAUDE.md Part II (COMMAND Patterns)

Bootstrap Patterns:

• manage-skills skill for skill creation standards
• manage-subagents skill for agent patterns
• manage-commands skill for command structure

Template Library:

• Consult existing templates in sys-core/skills/*/assets/templates/
• Follow progressive disclosure patterns
• Apply router pattern for complex skills