Compress raw document text into query-focused evidence snippets to reduce token usage
Single tool with critical definition gaps. Tool name 'compress_document_context' is verb-noun compliant, but input schema is present in description but NOT visible in the actual tool registration in mcp_server.py. The tools/list response shows only name and description, no input schema. Parameter descriptions exist in the docstring but are not formally declared in the MCP schema. Output schema is undocumented. Error handling is absent. The tool accepts parameters (query, raw_document_text) but the MCP registration does not expose them, forcing clients to guess at the interface.
Compress raw document text into query-focused evidence snippets.
Input schema not registered in MCP tools/list response. Tool accepts query and raw_document_text parameters but mcp_server.py does not declare them in the schema. Clients cannot discover parameter names, types, or descriptions.
Output schema undocumented. Tool returns a dict with query, original_word_count, compressed_word_count, compression_ratio, tokens_saved_estimate, retained_snippets, and compressed_context_text, but no schema is declared in the MCP response. LLMs cannot plan downstream operations or extract fields reliably.
No error handling or recovery guidance. Tool has no try-catch, validation, or error responses. If query or raw_document_text are missing or malformed, the tool silently uses defaults or crashes. No actionable error messages for LLMs.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 37 | 2026-07-28+ | v2 |
Parameter descriptions exist in client.py docstring but are not exposed in MCP schema. The mcp_server.py handle_mcp_request does not pass parameter metadata to clients, violating the pattern:tool-description requirement that every parameter must have a description.
Tool description is 79 characters, below the 10 - 1024 baseline and lacks WHEN/WHY context. Should explain when to use this tool vs alternatives, what prerequisites exist, and what the output enables downstream.