Server has 5 tools with complete schemas and descriptions, but quality is inconsistent. Tool names follow verb_noun convention (process_pdf, enrich_concepts, search_vault, get_note, vault_stats). Descriptions are present (avg ~120 chars) but lack actionable context for LLM selection. Parameters have types and descriptions, but several lack constraints (e.g., profile enum values not enforced, search_in lacks validation). No output schemas documented. Error handling is absent, no guidance on retryability, user-fixable vs fatal errors, or recovery paths. Security concern: process_pdf and enrich_concepts make external API calls (Claude API) but no mention of rate limiting, timeout handling, or permission gates. Composition is reasonable, tools are single-purpose and chainable (search_vault → get_note), but no pagination support for search results.

• Document output schema for each tool. Example for search_vault: 'Returns {results: [{note_name: string, path: string, excerpt: string}], total: integer, has_more: boolean}'.
• Add formal enum constraints to profile parameter: {type: 'string', enum: ['auto', 'philosophers_notes', 'shortform', 'readingraphics', 'getabstract', 'generic']}.
• Implement pagination for search_vault: add limit (default 20, max 100) and offset/cursor parameters; return total count and has_more flag.
• Add error handling: document which errors are retryable (e.g., 'API rate limit exceeded, retry in 60s') vs user-fixable (e.g., 'PDF not found at path, check vault/raw/ directory').
• Add timeout and rate-limit handling for process_pdf and enrich_concepts; document expected latency and max concurrent calls.