A FastAPI-based MCP server that provides tools for data processing, file compression, log sorting, and CSV analysis
This server has critical definition gaps across all four tools. Tool names lack action verbs (e.g., 'Parquet Reader' instead of 'read_parquet'). Descriptions are present but generic (10-50 chars, well below the 50-200 char LLM-optimized baseline). Input schemas are visible in the rubric summary but NOT in the source code, mcp_handlers.py shows only tool metadata dicts with 'id', 'name', 'description', 'usage' fields, with NO explicit JSON Schema definitions. The 'usage' field describes parameters informally ('file: filename (optional)') rather than as structured schemas. Output schemas are completely undocumented. Error handling is minimal, handlers return generic error dicts without recovery guidance. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk profiles (Compression Tool is WRITE). Parameters lack type constraints, enums, and validation rules. This server would not pass code review for production use.
Compresses files using gzip
Analyzes CSV files using pandas
Sorts log file entries by timestamp
Reads columns from Parquet files
No JSON Schema input definitions visible in source code. Tool metadata dicts contain only informal 'usage' strings, not structured schemas with types and constraints.
Tool names do not start with action verbs. 'Parquet Reader', 'Parallel Sorting', 'Compression Tool', 'Data Analysis using Pandas' should be 'read_parquet', 'sort_log_by_timestamp', 'compress_file', 'analyze_csv_data'.
Descriptions are too short (10-50 chars) and lack context for LLM selection. 'Reads columns from Parquet files' does not explain WHEN to use it, what it returns, or dependencies. Baseline is 50-200 chars for LLM-optimized descriptions.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 26 | 2026-07-28+ | v2 |
No output schemas documented. Handlers return dicts (e.g., compress_file returns status, original_file, compressed_file, sizes) but tool definitions do not declare these fields. LLMs cannot plan downstream calls without knowing response structure.
Compression Tool (WRITE risk) lacks destructiveHint annotation and confirmation pattern. Agents can compress files without warning. No dry-run or confirmation step to prevent accidental data loss.