An MCP server that helps users build legal case files by walking through facts, identifying evidence, and saving structured case data to Blueshoe's platform. Integrates with Claude agents to guide case documentation.
Two tools with basic schemas and descriptions, but significant gaps in parameter documentation and error handling. create_matter has a single string parameter with minimal context; add_matter_document has three string parameters but lacks guidance on valid document paths, content format, or error recovery. No output schemas documented. No error handling guidance. Descriptions are present but generic (10 - 20 chars above minimum). No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Both tools are WRITE operations but lack confirmation/dry-run patterns. Parameter descriptions do not explain constraints, formats, or dependencies.
Add a document to an existing matter (e.g., case-file/summary.md, case-file/timeline.md, case-file/evidence-checklist.md)
Create a new legal matter with a plain-language narrative description
No output schemas documented. LLMs cannot plan downstream calls or extract matter_id/document_id for chaining.
Parameter descriptions lack format/constraint guidance. 'narrative' does not explain length limits, required sections, or tone. 'document_path' does not list valid paths (case-file/summary.md, case-file/timeline.md, case-file/evidence-checklist.md are mentioned in system prompt but not in tool schema).
No error handling guidance. If create_matter fails (e.g., duplicate matter, auth failure), LLM has no recovery path. If add_matter_document fails (invalid path, matter not found), no actionable error message.
Destructive/write operations lack confirmation pattern. Both tools modify state (create matter, add document) but no dry-run, confirmation_required, or idempotent hints. Agents may execute without user approval.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 46 | 2026-07-28+ | v2 |
No tool annotations. Missing readOnlyHint, destructiveHint, idempotentHint. LLMs cannot infer safety properties or retry behavior.