container-security
Pass
Audited by Gen Agent Trust Hub on Jul 15, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill includes a dedicated 'Prompt Injection Safety Notice' that explicitly instructs the agent to treat all content within analyzed files (such as labels, annotations, or descriptions) as data rather than instructions. This effectively mitigates the risk of indirect prompt injection from untrusted configuration files.
- [DATA_EXFILTRATION]: Tool access is restricted to 'Read', 'Grep', and 'Glob'. The skill does not have access to network tools (curl, wget) or external communication capabilities, ensuring that any sensitive data discovered during the security review remains local.
- [EXTERNAL_DOWNLOADS]: There are no external dependencies, remote script executions, or package installations. All logic is contained within the provided markdown files.
- [COMMAND_EXECUTION]: The skill does not utilize shell execution, subprocess spawning, or dynamic code evaluation (eval/exec). It performs its analysis solely through pattern matching and manual inspection logic.
- [CREDENTIALS_UNSAFE]: While the skill contains regex patterns designed to detect hardcoded secrets (e.g., API keys, passwords) in the user's code, it does not contain any hardcoded credentials for its own operation and does not attempt to access system-level secrets.
Audit Metadata