brainstorm-first
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill requires the agent to read and process external context such as repositories, logs, and configuration files to inform the brainstorming phase.\n
- Ingestion points: Workflow instructions in Section 1 (SKILL.md) direct the agent to inspect repository artifacts, logs, and configuration files.\n
- Boundary markers: The skill does not specify the use of delimiters or instructions to ignore embedded commands within context data.\n
- Capability inventory: During the brainstorming phase, the skill explicitly prohibits implementation actions (writes, deployments, etc.), limiting activity to read-only discovery.\n
- Sanitization: No specific sanitization or filtering logic is defined for data retrieved from external sources.\n- [SAFE]: The skill implements significant behavioral guardrails that improve security. It includes 'Non-negotiable gates' that prevent the agent from taking actions or making assumptions without user consent. The mandatory 'Reality Review' (Section 3) explicitly requires testing for security, privacy, maintenance, and abuse potential before proposing solutions.
Audit Metadata