docker-agents-review

Pass

Audited by Gen Agent Trust Hub on Jun 30, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is a set of educational and auditing guidelines for Docker security. It provides structured checklists for identifying common vulnerabilities like running as root, hardcoding secrets, and using untrusted base images.
  • [PROMPT_INJECTION]: No evidence of prompt injection or behavior override patterns was detected. The instructions are clearly scoped to providing a framework for code review.
  • [DATA_EXFILTRATION]: No network exfiltration or unauthorized file access patterns were found. The skill correctly identifies hardcoded secrets as critical issues to be avoided in user configurations.
  • [REMOTE_CODE_EXECUTION]: The skill does not download or execute remote code. All referenced URLs point to official Docker documentation, which are considered well-known and trusted services.
  • [CREDENTIALS_UNSAFE]: While the skill contains strings that look like API keys (e.g., 'sk-123') and passwords, these are explicitly presented within 'BAD' examples to teach the agent what to flag during an audit. They are not functional credentials for the skill itself.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 30, 2026, 02:02 AM
Security Audit — agent-trust-hub — docker-agents-review