Two tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. save-memory has 5 parameters with descriptions; search-memory has 2. Both tools lack documented return schemas, making it unclear what fields agents should expect. Descriptions are present but generic (10-50 chars baseline). No error recovery guidance. No tool annotations (readOnlyHint/destructiveHint). Prompts are defined but add minimal value, they simply restate tool names. Overall structure is functional but falls short of production quality.

• Document return schemas for both tools. save-memory should return {success: boolean, path: string, sections_processed: number}. search-memory should return {results: [{content: string, relevance_score: number}], total_matches: number}.
• Add tool annotations: save-memory with destructiveHint=true (modifies vector DB), search-memory with readOnlyHint=true (no side effects).
• Expand tool descriptions to 100-150 chars. Example: 'Save content to vector database with semantic embeddings. Use this to store knowledge for later retrieval. Supports markdown, code, and plain text.'
• Add error recovery guidance. E.g., 'Error storing content: path already exists. Try a different path or use update-memory to replace.' Provide actionable next steps.
• Declare parameter constraints explicitly. For 'type', document valid values as enum: ['markdown', 'code', 'plaintext']. For 'maxMatches', specify range: 1-100 with default 10.