using-aegis
Pass
Audited by Gen Agent Trust Hub on May 29, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: No malicious patterns or bypass instructions were detected. The skill explicitly states that user and project instructions outrank its own rules, ensuring the user remains in control of the agent's behavior.
- [DATA_EXFILTRATION]: The skill lacks any code or instructions for exfiltrating data. It includes defensive privacy rules that instruct the agent to avoid reading sensitive files such as session history (
history.jsonl,~/.claude/projects) or large logs by default. - [COMMAND_EXECUTION]: The shell commands referenced in the skill are limited to benign git environment detection and workspace management (e.g.,
git rev-parse). These are standard for development tools and do not pose a security risk. - [REMOTE_CODE_EXECUTION]: No remote code execution patterns were found. The skill mentions a local workspace helper script executed via Python, but this is a path resolved from local environment variables or configuration, not a remote download.
- [SAFE]: The skill focuses on methodology and platform adaptation. It provides transparent documentation and tool mappings to ensure consistent and disciplined behavior across different AI CLI environments without introducing malicious capabilities.
Audit Metadata