| name | skill-creator |
| description | Create production-ready Agent Skills following the agentskills.io open standard. Use when a user asks to build, generate, or scaffold an agent skill with proper structure, security scanning (OWASP ASI01-10), and multi-platform packaging. |
| compatibility | Claude Code, OpenClaw, and agentic platforms supporting agentskills.io |
| metadata | {"generator":"agent-skill-creator-ts","version":"1.0.0","standard":"agentskills.io","template-types":"reactive, autonomous, workflow, suite, mcp-integration, knowledge"} |
Skill Creator
Create production-ready Agent Skills following the agentskills.io open standard with built-in security scanning and multi-platform packaging.
When to Use
Use this skill when:
- User asks to create, build, or generate an agent skill
- User wants to scaffold a new skill from a description
- User needs to validate an existing skill against the agentskills.io spec
- User wants to run a security scan on a skill (OWASP ASI01-10)
- User needs to package a skill for distribution with AI-BOM
How It Works
This skill uses an 8-phase generation pipeline:
- Discovery - Analyze user intent, detect APIs and tools needed
- Design - Determine capabilities, autonomy mode, use cases
- Architecture - Plan directory structure and token budgets
- Activation - Engineer description, generate YAML frontmatter
- Implementation - Generate SKILL.md body, scripts, references
- Security - Run 3-layer security scan (Static + Behavioral + LLM)
- Testing - Validate against agentskills.io spec
- Packaging - Write files, generate AI-BOM, create distributable package
Template Types
| Template | Use Case | Key Feature |
|---|
reactive | On-demand skills triggered by user | Scripts + error handling |
autonomous | Proactive skills with heartbeat/cron | HEARTBEAT.md + scheduling |
workflow | Multi-step with user checkpoints | Prepare-Propose-Build-Review-Deliver |
suite | Multiple related sub-skills | Shared references, orchestration |
mcp-integration | MCP server tool descriptions | Tool schemas + parameter tables |
knowledge | Pure knowledge/documentation | No scripts, markdown only |
Examples
Create a Reactive Skill
"Create a skill that formats code according to project conventions"
Create an Autonomous Skill
"Create a skill that monitors API endpoints and alerts when services go down"
Validate a Skill
"Validate the skill at ./my-skill against the agentskills.io spec"
Security Scan
"Run a security scan on ./my-skill and show the OWASP ASI coverage"
Package a Skill
"Package ./my-skill for distribution with AI-BOM"
Scripts
| Script | Purpose |
|---|
scripts/init-skill.ts | Generate a new skill from description |
scripts/validate-skill.ts | Validate against agentskills.io spec |
scripts/security-scan.ts | Run 3-layer security scan |
scripts/package-skill.ts | Package for distribution |
Error Handling
| Error | Cause | Resolution |
|---|
| Invalid template type | Unsupported template | Use one of: reactive, autonomous, workflow, suite, mcp-integration, knowledge |
| Frontmatter validation | Missing required fields | Ensure name and description are present |
| Security scan failure | Critical/high findings | Address security findings before packaging |
| Name mismatch | Frontmatter name != directory | Rename directory or update frontmatter name |
Anti-Patterns
What NOT to Do
- Do NOT skip the security scan before distributing skills
- Do NOT hardcode API keys or credentials in skill files
- Do NOT exceed 500 lines in SKILL.md (use references/ for overflow)
- Do NOT use eval(), exec(), or child_process in skill scripts
- Do NOT create skills that modify agent memory files (CLAUDE.md, MEMORY.md)
