Three tools with significant definition gaps. Tool names lack action verbs (createLearning is acceptable, but listLearnings and getCurrentWorkspace are weak). Descriptions are present but generic (10-50 chars, below the 50-200 char LLM-optimized baseline). Input schemas exist but are overly generic: 'query' and 'body' parameters use record(z.any()) with minimal type specificity. No output schemas documented. Parameters lack granular descriptions, 'Query string parameters' is not actionable. No error handling guidance. Tools appear dynamically generated from OpenAPI spec, which limits schema precision and parameter validation.

• Rename tools to use consistent verb_noun pattern: 'create_learning', 'list_learnings', 'get_workspace'. Verify names match REST semantics.
• Expand descriptions to 80-150 chars. Example: 'createLearning: Store a durable learning record scoped by product. Use this to persist agent insights for recall in future sessions. Returns the created learning ID and timestamp.'
• Replace generic 'record(z.any())' schemas with explicit field definitions. For createLearning.body, define: product (required, string, 'stable product identifier'), category (required, enum), title (required, string, max 200), body (required, string), confidence (required, enum: low|medium|high).
• Document output schemas for each tool. Example: 'listLearnings returns {learnings: [{id, product, category, title, body, confidence, created_at}], total: number, next_cursor?: string}'.
• Add error handling guidance. Example: 'If product not found, returns 404 with message. Try create_learning first to initialize the product scope.' Implement recovery hints per pattern:recovery-guide.