Single tool with clear naming (get_lifecycle_evidence) and documented input schema. Description is adequate (71 chars) but lacks WHEN/WHY guidance. Input parameters have types and descriptions, but scope parameter is underspecified (generic dict). Output schema (GetLifecycleEvidenceOutput) is not visible in provided code, only inferred from return type annotation. No error handling guidance, no examples of failure modes, no pagination despite potential for large result sets. Follows basic tool structure but lacks production-grade polish.

• Document GetLifecycleEvidenceOutput schema explicitly in code comments or docstring. Include field names, types, and descriptions. Example: 'Returns {records: List[LifecycleRecord], total_count: int, next_cursor: str|null}'.
• Replace generic scope: dict[str, str] with a concrete Pydantic model or TypedDict. Define valid scope keys (e.g., 'region', 'department', 'timestamp_range') with descriptions and constraints.
• Add pagination: accept limit (1-100, default 20) and cursor parameters. Return total_count and next_cursor in response. Document in tool description: 'Results are paginated; use cursor for subsequent pages.'
• Expand tool description to 100-150 chars. Add: 'Retrieve historical evidence records for an entity attribute within a scope. Use this to audit changes, verify compliance, or trace data lineage. Requires entity name, attribute name, and scope filters.'
• Add error handling documentation. Describe expected failures: 'Returns error if entity not found (suggest search_entities first), if scope keys are invalid (list valid keys), or if query times out (retry with narrower scope).'

Inferred effective spec: 2026-07-28+.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry