MCP Server for extracting and cleaning structured JSON from noisy LLM outputs by removing markdown code fences and isolating structured payloads.
Single tool with minimal definition quality. Tool name 'extract_clean_json' lacks action verb clarity (should be 'extract_json' or 'parse_json'). Description is adequate (73 chars) but lacks WHEN/WHY guidance. Input schema present but parameter 'raw_text' has no description, violates pattern:tool-description. No output schema documented. No error handling guidance. Tool does one thing well but definition falls short of production standards.
Strip markdown code blocks and extract structured JSON payload.
Parameter 'raw_text' lacks description in schema. LLMs cannot infer what format is expected (plain JSON, markdown fence, mixed narrative).
No output schema documented. Tool returns JSON object but LLMs don't know structure, field names, or types of extracted payload.
Error handling provides no recovery guidance. ValueError 'No valid JSON payload could be extracted' gives LLM no actionable next step.
Tool name 'extract_clean_json' is redundant ('extract' + 'clean' + 'json'). Simpler name 'extract_json' or 'parse_json' is clearer.
Description lacks WHEN/WHY context. Does not explain when to call this vs. direct JSON parsing, or what formats it handles.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 42 | 2026-07-28+ | v2 |