Three tools with severe definition gaps. All lack proper input schemas (no type definitions visible), descriptions are vague or missing parameter details, and no output schemas documented. Tool names lack action verbs (digest-runner, build_client_guide). No error handling guidance. Descriptions range 50-200 chars but fail to explain WHEN to use each tool or what it returns. Parameters exist but lack type constraints and validation rules. This server reads as a collection of scripts, not an MCP tool interface.

• Add proper JSON Schema to all tools: define 'type' (string, array, object) for each parameter. For 'skill' in digest-runner, use enum: ['meetings-digest', 'weekly-rollup'] instead of inline description.
• Rename tools with action verbs: 'digest-runner' → 'run_digest', 'build_client_guide' → 'generate_client_guide_docx', 'build_client_guide_html' → 'generate_client_guide_html'.
• Document output schemas. E.g., digest-runner should return {log_path: string, exit_code: number, skill_executed: string}. build_client_guide should return {output_path: string, file_size_bytes: number}.
• Add parameter constraints: 'out' param should specify format (file path, max length 255 chars, must be writable directory). 'skill' should use enum constraint, not free-form string.
• Add error handling guidance: 'If Plaud OAuth fails, check PLAUD_API_KEY env var. If OneDrive is unreachable, verify network and retry.' Include retryable vs fatal error classification.