Agentify exposes 3 tools with partial schema coverage. All tools have descriptions (10-50 chars), but descriptions are generic and lack actionable context for LLM selection. Input schemas are present but incomplete: 'transform' and 'self-describe' lack type definitions for most parameters; 'parse' has better typing but still missing descriptions for some fields. No output schemas documented. No error handling guidance. Tools combine multiple concerns (e.g., 'transform' handles parsing, generation, and file I/O in one call). Parameter naming is inconsistent: 'format' is an array but lacks enum constraints; 'output' appears in multiple tools but lacks validation rules. No evidence of idempotency, pagination, or recovery patterns.

• Add explicit output schemas for all tools. Document what fields parse() returns in the IR, what files transform() generates, and what self-describe() outputs.
• Split 'transform' into separate tools: parse_openapi_spec (returns IR), generate_mcp_server (IR → MCP project), generate_claude_md (IR → markdown), etc. Each tool should do one thing.
• Add enum constraint to 'format' parameter: {"type": "array", "items": {"type": "string", "enum": ["mcp", "claude.md", "agents.md", "cursorrules", "llms.txt", "gemini.md", "skills", "a2a", "cli"]}}
• Expand descriptions to 100-200 chars. Example: 'Parse an OpenAPI spec into an intermediate representation (IR). Use this to inspect API structure before generating agent interfaces. Returns domain names, capabilities, and auth requirements.'
• Add error handling: catch parse failures and return actionable messages like 'Invalid OpenAPI spec: missing required field "paths". Ensure spec is valid JSON/YAML and includes endpoint definitions.'