ask-first-act-later
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [PROMPT_INJECTION]: The skill uses authoritative language such as 'Overrides the agent's behavior' and 'must immediately halt' to redirect the agent's default operational mode. These patterns are used to implement a legitimate behavioral shift into a 'consultant mode' rather than to bypass safety guardrails.
- [INDIRECT_PROMPT_INJECTION]: The skill acts on user-provided requests to generate clarifying questions, creating a potential surface for indirect injection.
- Ingestion points: User requests and refinement answers processed in
SKILL.md(Protocol steps 1-4). - Boundary markers: Absent; the instructions do not specify delimiters for separating user-provided data from the agent's internal logic.
- Capability inventory: None; the skill is limited to natural language interaction and lacks file system, network, or command execution capabilities.
- Sanitization: Absent.
Audit Metadata