Loading ToolBench
AWS Model Context Protocol Server - A lightweight service that enables AI assistants to execute AWS CLI commands through the Model Context Protocol (MCP).
Two tools with basic schemas and descriptions, but significant gaps in parameter documentation, error handling guidance, and output schema clarity. aws_cli_help has a minimal description (89 chars); aws_cli_pipeline is better (234 chars) but lacks actionable error recovery guidance. Neither tool documents output structure or pagination. The 'command' parameter in aws_cli_pipeline accepts arbitrary strings with pipes, creating injection risk without validation hints in the description. No tool annotations (readOnlyHint/destructiveHint) despite clear risk profiles. Parameter descriptions are sparse, 'command' in aws_cli_pipeline lacks format/constraint guidance.
Get AWS CLI command documentation. Retrieves the help documentation for a specified AWS service or command by executing the 'aws <service> [command] help' command.
Execute an AWS CLI command, optionally with Unix command pipes. Validates, executes, and processes the results of an AWS CLI command, handling errors and formatting the output for better readability. The command can include Unix pipes (|) to filter or transform the output, similar to a regular shell. The first command must be an AWS CLI command, and subsequent piped commands must be basic Unix utilities.
aws_cli_help description is 89 characters, below the 10 - 1024 baseline and lacks WHEN/WHY context. Does not explain when to call this vs aws_cli_pipeline or what output structure to expect.
aws_cli_pipeline 'command' parameter has no format constraints, length limits, or injection-prevention guidance in description. Accepts arbitrary pipes and Unix commands without documenting validation rules or safe/unsafe patterns.
No output schemas documented for either tool. LLMs cannot infer what fields to expect from aws_cli_help (JSON? text?) or aws_cli_pipeline (exit code? stdout? stderr? combined?). Breaks downstream tool chaining.
No tool annotations (readOnlyHint/destructiveHint/idempotentHint) despite clear risk profiles: aws_cli_help is READ_ONLY, aws_cli_pipeline is WRITE. Missing annotations prevent agents from reasoning about safety and retry semantics.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 47 | <=2025-11-25 | v2 |
Error handling lacks recovery guidance. No description of what errors are retryable, what indicates user misconfiguration, or how to self-correct. E.g., 'Invalid AWS CLI command' tells agent nothing; should be 'Invalid command: got "s3 invalid-op". Valid s3 commands: ls, cp, sync, rm. Try aws_cli_help s3 first.'