Weave Compose has 5 tools with basic schemas and descriptions, but significant gaps limit production readiness. All tools have descriptions (10-80 chars, mostly adequate) and input schemas with type definitions. However, output schemas are completely undocumented, LLMs cannot plan downstream calls or extract structured data. Parameter descriptions are minimal (e.g., 'Path to directory containing skill files' lacks format/validation guidance). No error handling guidance, no recovery paths, and no indication of which operations are destructive vs. safe. Tool naming follows verb_noun convention (get_, post_*) but post_* is non-standard for MCP (should be create_, update_, etc.). The skill composition domain is niche but the implementation lacks production-grade polish.

• Rename post_load → load_skills, post_query → search_skills, post_compose → compose_skills to follow verb_noun MCP convention and improve LLM intent inference.
• Document output schemas for all tools. E.g., get_skills should return {skills: [{id, name, platform, capabilities_count}], total_count}. Include field types and descriptions.
• Add validation constraints to parameters: path must be absolute or relative, confidence_threshold must be 0.0 - 1.0, top_n must be 1 - 50. State these in descriptions.
• Add error handling guidance: 'If path does not exist, verify the directory path and try again. If permission denied, check file permissions.' Include recovery paths.
• Mark post_load as destructive in description: 'Overwrites the current session. Call get_status first to review loaded skills before replacing.' Consider adding a dry_run parameter.