skills/a-canary/arc-skills/overseer/Gen Agent Trust Hub

overseer

Fail

Audited by Gen Agent Trust Hub on Sep 20, 2026

Risk Level: HIGHPERSISTENCECOMMAND_EXECUTION
Full Analysis
  • [PERSISTENCE]: The skill's instructions in SKILL.md explicitly mandate the configuration of a cron job (*/15 * * * * bash <skill>/overseer.sh) to ensure the monitoring script runs in the background every 15 minutes. This creates a persistent execution loop that remains active across user sessions.
  • [COMMAND_EXECUTION]: The overseer.sh script utilizes the herdr pane send-keys command to inject keystrokes (such as the 'enter' key) into other active terminal panes. This allows the skill to automate interactions with separate processes and terminal sessions without direct user review or authorization.
  • [DATA_EXPOSURE]: The skill captures the last 50 lines of output from every accessible terminal pane using herdr pane read. This provides the automated script with visibility into potentially sensitive data, credentials, or private information displayed in other active sessions on the system.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Sep 20, 2026, 08:53 PM
Security Audit — agent-trust-hub — overseer