Asset-Aware MCP has 4 tools with partial schema and description coverage. Two tools (ingest_docx, get_docx_content) have detailed descriptions in Chinese/English mix, but schemas are minimally specified, parameter types are present but lack comprehensive validation constraints. The other two tools (get_job_status, list_jobs) have generic, minimal descriptions. No tools declare output schemas or error handling guidance. Parameter descriptions are sparse; input constraints (min/max, enums) are not enforced. No tool declares permissions, idempotency guarantees, or whether calls are retryable. The server's STDIO-only transport limits remote accessibility, but definition quality issues persist independently.

• Document output schemas for all 4 tools. Specify the return type (object, array, string), top-level fields, and field types. Example: 'Returns {doc_id: string, output_dir: string, parts_count: integer, assets_count: integer}'.
• Expand tool descriptions to 50-200 characters. Explain WHAT the tool does, WHEN to call it (e.g., 'Call after uploading a file to parse and structure it'), and WHAT it returns. Example for list_jobs: 'List all ETL ingestion jobs. Returns active and completed jobs with status, progress, and result summary. Use this to track async ingest_docx operations.'
• Add explicit input constraints. For ingest_docx file_path: 'Absolute or relative path to a .docx or .doc file. Must exist and be readable. Maximum file size: [X] MB.' For get_docx_content max_chars: 'Optional. If provided, must be a positive integer (1 - 1000000). Omit for full content.'
• Implement and document error handling. For ingest_docx: 'Returns error if file not found, not readable, or parsing fails. Retryable on timeout; user-fixable on path/permission errors. Consider calling search_documents() if unsure of file location.' For get_job_status: 'Returns 404 if job_id is invalid or expired. Retryable until job completes.'
• Add block_id format documentation to get_docx_content. Clarify: 'block_id must match one of the blocks returned by ingest_docx (e.g., p001 for paragraph 1, t001 for table 1, h001 for heading 1). Omit to retrieve entire document.'
• Add pagination to list_jobs. Include offset/limit or cursor parameters with documented defaults. Example: 'limit (1 - 100, default 20): max jobs to return. offset (default 0): starting position. Returns {jobs: [...], total: integer, has_more: boolean}'.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry