crack
Pass
Audited by Gen Agent Trust Hub on Oct 19, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements robust security controls across its entire stack. Key findings include:
- Environment Scrubbing: The orchestration scripts explicitly remove sensitive environment variables (e.g., ANTHROPIC_, CLAUDE_CODE_, OPENAI_*) before spawning sub-processes, preventing unintentional credential leakage to child agents.
- Path Security: Extensive use of
refuseSymlinksandcanonicalPathSyncprevents directory traversal attacks and ensures operations remain within approved workspace boundaries. The log viewer usesO_NOFOLLOWwhen opening files to prevent symlink-based attacks. - Viewer Authentication: The activity viewer binds only to
127.0.0.1and requires a randomly generated 24-character hex token for all API access, ensuring that local data is not exposed to other users or network entities. - Request Validation: The orchestration protocol uses a strict JSON schema with a 'forbidden keys' list (including
command,exec,shell) to ensure that delegated requests describe work rather than carrying executable code. - Data Redaction: The viewer adapter includes automated redaction of credentials, local file paths, and email addresses from the logs before they are displayed or exported.
Audit Metadata