handoff-that-to-agent-pane

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The skill intentionally instructs launching autonomous agents with flags that skip permissions and disable sandboxes while handing them absolute conversation snapshots and the current working directory, creating a high risk of data exfiltration and enabling remote code execution.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). Required runtime workflow ingests the user-authored request to build a “brief” and then sends it (including the outsider-provided assignment/context) to a new Herdr agent via herdr agent prompt <name> "<text>", after reading a snapshot path.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.70). This skill explicitly instructs launching agents with flags that disable sandboxes or skip permission prompts (e.g. --sandbox disabled, --dangerously-skip-permissions, --approve, --approve-for-me, --yolo), which encourages bypassing local security safeguards even though it does not directly request sudo or create users.