docker-swarm
Pass
Audited by Gen Agent Trust Hub on Nov 22, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill enforces a strict 'read-only' policy for cluster reviews, using diagnostic commands like
docker node lsanddocker service inspectto gather evidence without altering the environment. - [SAFE]: Comprehensive instructions for redacting secrets (e.g., Kind and location only, never values) are provided to ensure sensitive data is not exposed during the audit process.
- [SAFE]: The skill includes explicit rules of engagement that instruct the agent to treat audited content (e.g., stack-file comments, scripts) as evidence to be reported rather than instructions to be followed, effectively mitigating indirect prompt injection vectors.
- [SAFE]: All referenced technical principles are grounded in established industry literature (e.g., O'Reilly and Manning publications), and the skill does not utilize external scripts, obfuscation, or unverifiable dependencies.
Audit Metadata