Promptly has 2 tools with clear naming (verb_noun pattern) and detailed descriptions. refine_prompt includes comprehensive guidance on WHEN/SKIP FOR and monorepo handling. However, schema completeness is mixed: refine_prompt has full Zod schemas with descriptions, but get_refinement_rules lacks input parameter descriptions. Output schemas are not documented, LLMs cannot predict return structure. Error handling is absent; no recovery guidance or categorization. Parameter descriptions are strong (target_files, context_files explain monorepo routing), but missing output documentation and error cases significantly limit production readiness.

• Document output schema for refine_prompt: return type should specify { refined_prompt: string, intent: string, detected_files: string[], analysis_root: string, ... }. This enables LLMs to extract and use the refined prompt correctly.
• Add error handling to refine_prompt description: 'If project_path is invalid, returns error with suggestion to verify path. If git is unavailable, analysis proceeds without recent-file hints. If codebase analysis times out (>30s), returns partial context with available metadata.'
• Add description to get_refinement_rules 'agent' parameter: 'Target agent for rule formatting. claude_code: Claude Code Editor rules; cursor: Cursor IDE rules; gemini_cli: Google Gemini CLI rules; qwen_code: Alibaba Qwen Code rules; generic: agent-agnostic rules.'
• Validate project_path input: document that it must be an absolute path to a valid directory. Return clear error if path doesn't exist: 'project_path must be an absolute path to an existing directory. Got: <path>. Verify the path and try again.'
• Declare refine_prompt as idempotent in description: 'Calling refine_prompt twice with the same inputs returns the same refined prompt. Safe to retry on transient failures.'