hermes-gateway-doctor
Pass
Audited by Gen Agent Trust Hub on Jul 25, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a robust 'Untrusted Content Boundary' defined in
SKILL.mdandreferences/safety.md. This section explicitly instructs the agent to treat all inspected content (logs, messages, files) as untrusted evidence rather than instructions, effectively mitigating indirect prompt injection risks. - [SAFE]: The diagnostic workflow is designed to be read-only by default. The 'Safety Contract' and 'Approval Boundary' require separate, explicit user approval before the agent can perform any mutations, such as process restarts or state changes.
- [SAFE]: The skill prioritizes credential privacy by prohibiting the printing or logging of tokens. It uses a validation strategy that only reports whether required credentials are 'set' or 'not set' without revealing the values.
- [SAFE]: The included validation and test scripts (
scripts/validate_bundle.pyandtests/test_contracts.py) use only the Python standard library, minimizing supply chain risks and avoiding unverifiable third-party dependencies.
Audit Metadata