team-brainstorm
Pass
Audited by Gen Agent Trust Hub on May 5, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface due to the collaborative nature of its agent roles.
- Ingestion points: The coordinator and specialized worker roles (ideator, challenger, synthesizer, evaluator) read and process markdown files from the session directory that are created and updated by other agents.
- Boundary markers: The skill lacks explicit instructions or delimiters to isolate the text being processed, making it potentially vulnerable to instructions embedded within the generated ideas or critiques.
- Capability inventory: The agents have access to powerful tools, including
Bashfor shell access,Write/Editfor file modification, and theAgenttool for spawning sub-agents. - Sanitization: There is no logic for sanitizing or validating the content of the generated artifacts before they are ingested by subsequent agents.
- [COMMAND_EXECUTION]: The coordinator role uses the
Bashtool to run thepwdcommand. This is used to programmatically determine the project and skill root directories for path resolution during session initialization.
Audit Metadata