PeopleMesh defines 3 tools with clear, action-oriented names (get_, search_, match_) and detailed descriptions (150-250 chars each). All tools have input schemas with typed parameters and descriptions. However, output schemas are not documented in the provided code, and error handling guidance is absent. Tool descriptions are strong on usage intent but lack explicit recovery paths for common failures. Parameter descriptions are good but could specify constraints (e.g., max query length of 500 chars is mentioned but not formalized as a schema constraint).
Retrieve the authenticated user's PeopleMesh profile as JSON. Use this when profile data is explicitly needed. Treat missing fields as unknown and do not infer facts that are not present in the profile.
Find matches using the authenticated user's existing PeopleMesh profile as the source. Use this for requests like 'match my profile' or 'find opportunities for me'. Do not use it as a substitute for a new natural-language search request. Apply optional type and country filters only when the user explicitly requested them.
Search PeopleMesh from a natural-language query using the server-side query parser and ranking logic. Provide the user's request verbatim in query. Do not rewrite, broaden, relax, or reinterpret the criteria unless the user explicitly asks. Never set country automatically from locale, profile, session, or prior context; leave it empty unless the user explicitly requested a country or location filter. Use type only when the user clearly requested a result category such as PEOPLE, JOB, COMMUNITY, EVENT, PROJECT, or INTEREST_GROUP. Returns JSON with both the parsedQuery and the ranked results.
Output schemas not documented. LLMs cannot plan downstream calls or extract required fields without knowing what each tool returns.
No error handling guidance. Descriptions lack recovery hints (e.g., 'If query is too broad, try adding type or country filters'). Agents cannot self-correct on failures.
Parameter constraints not formalized in schema. Query max length (500 chars) is mentioned in description text but not enforced via JSON Schema maxLength. LLMs may ignore text constraints.
Enum values for 'type' parameter not formally declared. Description lists valid categories (PEOPLE, JOB, COMMUNITY, EVENT, PROJECT, INTEREST_GROUP) but schema should use enum constraint.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 68 | <=2025-11-25 | v2 |
No pagination or result limit documented. If match_prompt returns hundreds of results, context window bloat and token waste are likely. Schema should include limit/offset params and cap results.