Server defines 4 tools with basic schemas and descriptions, but significant gaps limit production readiness. Tool names lack clear action verbs (config_origin_inspect_inline vs inspect_config_origin_inline). Descriptions are present but minimal (10-60 chars, below the 50-200 char baseline for LLM optimization). Parameter descriptions are absent or trivial, recordingJson and recordingPath lack guidance on format, constraints, or error cases. Output schemas are undocumented; tools return JSON but no schema is declared for downstream tool chaining. Error handling is minimal, no recovery guidance or actionable error messages.

• Rename tools to start with action verbs: 'inspect_origin_recording_inline', 'verify_origin_recording_inline', 'inspect_origin_recording', 'verify_origin_recording'. This makes intent clear before the LLM reads the description.
• Add parameter descriptions for all inputs. For recordingJson: 'JSON string containing the origin recording (max 4194304 bytes). Must be valid JSON with keys: [list expected keys]. On parse error, returns ConfigOriginProofError with code INVALID_JSON.' For recordingPath: 'Workspace-relative path to the recording file (e.g., recordings/config.json). Must exist and be readable. Returns error if file not found or exceeds 4MB.'
• Document output schemas for all tools. E.g., 'Returns object with keys: {status: string, findings: array, digest: string, ...}. See examples/output.json for sample structure.' Include this in tool descriptions or a separate schema field.
• Add explicit input validation and error messages. Instead of generic errors, return: 'Invalid recordingJson: expected object with keys [x, y, z], got [a, b]. See README.md#recording-format for schema.' This guides LLM self-correction.
• Expose output schemas in the tool definition. Modify index.js tool() helper to include an outputSchema field, and update mcp-server.mjs to declare result schemas in tools/list response. This enables LLM planning and downstream tool chaining.