Credential vault for AI agents — store, retrieve, and manage API keys securely
Three tools with basic schemas and descriptions, but significant gaps in parameter documentation and output structure. All tools are READ_ONLY and stateless, which is good. However, parameter descriptions are minimal or absent (e.g., 'cost' parameter lacks context on what 'cost' means or how it affects behavior). No output schemas documented. Error handling returns plain strings without guidance on recovery. Tool names follow verb_noun pattern (list_, get_, vault_) which is correct, but descriptions are brief (34-72 chars) and lack context on when to use each tool vs. alternatives.
Retrieve a decrypted credential value from the vault (TTL 300s). Optional `cost` charges this agent's budget.
List all credential names available to this agent.
Check vault connection status and version.
Parameter 'cost' in get_credential lacks description of its purpose, units, and side effects. LLM cannot determine when to include it or what value to pass.
No output schemas documented for any tool. Responses are plain strings; LLM cannot parse structured data or chain results to downstream tools.
Error responses are unstructured strings (e.g., 'Error: 404, Not found'). No recovery guidance or error classification. LLM cannot determine if error is retryable or user-fixable.
Tool descriptions are too brief (34-72 chars, below 50-200 char baseline). 'Check vault connection status and version' does not explain when to call this vs. other tools or what the response contains.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 53 | 2026-07-28+ | v2 |
get_credential returns raw credential value as plain string. No metadata (TTL, access timestamp, cost charged). LLM cannot verify success or track usage.