Extracts structured values from messy product specifications and enriches product listings with parsed attributes like weight and packaging type.
This server has a single tool 'enrich_listing' with a basic implementation. The tool is defined only in example_usage.py and client.py as a Python class method, not as an explicit MCP tool registration. No MCP server framework is visible, this appears to be a standalone Python SDK without MCP protocol implementation. The tool has a short description (77 chars, below baseline of 194), parameters lack detailed constraints, no output schema is documented, no error handling guidance is provided, and there is no evidence of MCP transport layer. The implementation is functional as a Python SDK but does not meet MCP server quality standards.
Extracts structured values from messy specs and enriches product listings with parsed attributes.
Tool registration and MCP protocol implementation missing. Code shows a Python class method, not an MCP-registered tool. No evidence of MCP transport, tool definition messages, or protocol compliance.
No documented output schema. The enrich_listing method returns a dict with 'enriched_product_details' nested object, but LLMs have no formal schema specifying the structure, types, or field descriptions of the response.
Parameter descriptions lack constraints and format guidance. 'unstructured_specs' has a generic description with no hints about expected format, length limits, or examples of valid input. 'product_name' lacks guidance on required format.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 30 | 2026-07-28+ | v2 |
Tool description is too short (77 characters) and lacks WHEN/WHY guidance. At 77 chars (below p10=34 is acceptable but below the effective baseline of 194), the description 'Extracts structured values from messy specs and enriches product listings with parsed attributes.' does not explain when to call this vs other tools, what prerequisites exist, or what scenarios it handles.
No error handling or recovery guidance. The implementation silently defaults to weight=1.0 on parse failure and packaging='Standard Box' on missing keywords. LLMs receive no indication of what went wrong, whether the extraction succeeded, or what confidence to place in the results.
Hard-coded heuristics with no validation. The extraction logic uses fragile string matching ('lbs', 'eco', 'biodegradable') and floating-point parsing with bare try/except. No validation of numeric ranges, no bounds on weight values, no enum for packaging types.
Output always reports extraction_confidence=0.95 without justification. Confidence should reflect actual parse success (e.g., lower if parsing failed, higher only when pattern matches are certain). This misleads LLMs into trusting unreliable extractions.