agency-incident-responder
Pass
Audited by Gen Agent Trust Hub on Jul 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides detailed PowerShell and Bash scripts intended for manual execution by a system administrator. These scripts perform wide-ranging data collection, including processes, network connections, and persistence mechanisms.\n- [DATA_EXFILTRATION]: The triage scripts aggregate sensitive system data (including Security logs, SSH authorized keys, and shell profiles) into local directories. While the data remains local, the Linux script creates files in /tmp without explicit restrictive permissions, which could temporarily expose sensitive information to other local users. No automated network exfiltration was detected.\n- [PROMPT_INJECTION]: The instructions use strong imperative language to enforce a methodical persona. These instructions align with safety and forensic best practices (e.g., 'never modify evidence') rather than attempting to bypass safety filters.
Audit Metadata