Three tools are registered with FastMCP and have visible input schemas. All tools have descriptions (10-50 chars), though they are quite brief and lack actionable details. Input schemas are present with type and description annotations, meeting basic requirements. However, descriptions fall short of the 50-200 char optimal range for LLM optimization, and lack context on WHEN to use each tool, dependencies, and error recovery guidance. No output schemas are documented. Tools are READ_ONLY but lack explicit idempotentHint annotations.
Get detailed information about a specific EC2 instance
List EC2 instances in the current region
List S3 buckets in the current region
Descriptions are extremely brief (20-48 chars) and lack LLM-optimization guidance. They do not answer WHAT, WHEN, or WHY. E.g., 'List EC2 instances in the current region' does not explain when to call this vs describe_ec2_instance, what structure is returned, or how to filter results effectively.
No documented output schemas. Tools return strings (per code: 'async def list_ec2_instances(region: str, state: str = "all") -> str'), but LLMs cannot infer what fields or structure the response contains. This prevents downstream tool composition and forces the LLM to parse unstructured output.
Parameter descriptions are minimal. 'region' is described as 'AWS region to query' but lacks guidance on format (e.g., 'us-east-1') or how default region selection works. 'state' enum values are listed but not explained (e.g., what does 'pending' mean vs 'running'?).
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 56 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 40 | - | v1 |
No error handling guidance. Tools may fail (invalid region, missing credentials, permission denied) but descriptions provide no recovery hints like 'If region is invalid, call list_ec2_instances with region="us-east-1" first.' or 'Requires AWS credentials with EC2 read permissions.'
No result limit or pagination guidance. list_ec2_instances and list_s3_buckets could return large result sets, yet no description mentions limits, pagination support, or context-window implications.
Missing tool annotations (toolAnnotations=false). These are READ_ONLY tools but lack explicit readOnlyHint or idempotentHint annotations in the tool schema, which would clarify their safe nature to LLMs.