roast
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill demonstrates a vulnerability to indirect prompt injection by design, as it processes untrusted user input and passes it to multiple subagents.
- Ingestion points: The agent ingests user-provided ideas via the
$ARGUMENTSvariable and follows up with clarifying questions to build a context brief (SKILL.md). - Boundary markers: User-provided content in the 'brief' is directly interpolated into subagent system instructions (e.g., 'THE BRIEF: [brief]') without explicit delimiters or instructions to ignore embedded commands (SKILL.md).
- Capability inventory: The skill uses
general-purposesubagents and grants the 'Researcher' persona access to web search tools. It does not appear to use more sensitive tools such as file-system access or shell execution (SKILL.md). - Sanitization: There is no logic for sanitizing or validating the content of the user's idea before it is processed by the council of personas (SKILL.md).
Audit Metadata