AgentCore MCP Gateway with Stytch B2B OAuth authentication and scope-based access control for MCP tools
Both tools have minimal descriptions (under 30 chars) and lack parameter descriptions entirely. Input schemas are present but parameters lack type information beyond 'string'. No output schemas documented. Tool names follow verb_noun convention but descriptions fail to explain WHEN to use each tool or what they return. No error handling guidance. This is a proof-of-concept implementation, not production-ready.
Get current time for a timezone
Get weather for a location
Tool descriptions are critically short (under 30 chars). 'Get weather for a location' and 'Get current time for a timezone' lack context for LLM selection. No explanation of return format, when to prefer one tool over another, or error cases.
Parameter 'location' in get_weather lacks description. Parameter 'timezone' in get_time lacks description. LLMs cannot infer that 'timezone' expects IANA format without explicit guidance. Rubric requires every parameter to have a non-empty description.
No output schemas documented. LLMs do not know what fields to expect from these tools (e.g., does get_weather return {temp, condition, humidity} or a free-text string?). Rubric baseline: 100% of A+ tools have documented return types.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 32 | <=2025-11-25 | v2 |
No error handling guidance. If get_weather receives an invalid location or get_time receives an invalid timezone, what does the tool return? How should the LLM recover? Rubric requires error responses to guide the agent on next steps.
Tool definitions appear to be inferred from CDK stack code rather than explicitly registered with full schemas. Source shows tool names and basic input schemas but no explicit tool registration handler or output schema definitions visible in the provided code.