mid-engagement-ir-detection

Pass

Audited by Gen Agent Trust Hub on Jul 31, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill includes shell script snippets for automating engagement monitoring tasks.
  • Evidence: Use of curl, bc, and python3 heredocs in the 'Tooling — automated state-change detection' section to measure response times and process JSON logs.
  • [EXTERNAL_DOWNLOADS]: The provided automation logic performs network requests to target infrastructure.
  • Evidence: The bash watcher script utilizes curl -sk to fetch timing data from user-supplied target URLs.
  • [DATA_EXFILTRATION]: The skill instructs the operator to capture and persist engagement-sensitive data to local files.
  • Evidence: Instructions describe saving response headers, WAF cookies, and authentication state (from o365_attempts.json) into local files such as engagement_log/baseline.json and engagement_log/journal.jsonl.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 31, 2026, 04:47 PM
Security Audit — agent-trust-hub — mid-engagement-ir-detection