anthropic-evals-and-observability
Pass
Audited by Gen Agent Trust Hub on Jun 17, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill consists of markdown documentation and templates designed to guide developers in building robust evaluation pipelines. It does not contain executable scripts or automation that performs sensitive operations.
- [DATA_EXPOSURE]: The skill actively promotes security best practices by instructing users to never commit API keys to source control and to redact PII and secrets from evaluation datasets, traces, and logs.
- [PROMPT_INJECTION]: No evidence of prompt injection, instruction overrides, or safety filter bypasses was found. The skill focuses on measurement and monitoring rather than model interaction manipulation.
- [REMOTE_CODE_EXECUTION]: The skill does not perform any remote code downloads or executions. The code snippets provided in the templates are for illustrative purposes for the developer and are not executed by the agent.
Audit Metadata