use-goal
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user objectives which are executed autonomously by the agent, creating a surface for indirect prompt injection. 1. Ingestion points: objectives passed to /use-goal (SKILL.md). 2. Boundary markers: objective is stored in current.md under a Markdown header without additional escaping. 3. Capability inventory: agent is instructed to work autonomously, diagnose and fix failures, and not stop until completion, increasing the potential impact of malicious objectives. 4. Sanitization: the skill explicitly forbids bypassing security, consent, or destructive-action requirements in its stop conditions.
Audit Metadata