executing-plans
Pass
Audited by Gen Agent Trust Hub on Feb 20, 2026
Risk Level: SAFE
Full Analysis
- [SAFE] (SAFE): The skill consists of natural language instructions for task management. No malicious patterns, obfuscation, or unauthorized access attempts were detected.
- [Indirect Prompt Injection] (LOW): The skill is designed to ingest and act upon externally provided 'plan files'.
- Ingestion points: External implementation plans are read in 'Step 1: Load and Review Plan'.
- Boundary markers: Absent in this specific file; the skill relies on the agent's critical review to identify concerns.
- Capability inventory: The agent is authorized to execute tasks in batches, which may involve tool usage depending on the plan content.
- Sanitization: The skill mandates 'review checkpoints', 'human partner' consultation for concerns, and explicit 'STOP' triggers for blockers or unclear instructions.
Audit Metadata