brainstorming
Pass
Audited by Gen Agent Trust Hub on Jul 6, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns or security vulnerabilities were detected. The skill follows best practices for collaborative design and agent safety.
- [PROMPT_INJECTION]: The instructions do not contain markers intended to bypass safety filters, extract system prompts, or override agent constraints. The use of a
<HARD-GATE>section is an internal safety mechanism to ensure user review. - [DATA_EXFILTRATION]: The skill does not access sensitive files (e.g., credentials, SSH keys) or utilize network tools to send data to external domains. Interaction with project files is limited to understanding context for design.
- [REMOTE_CODE_EXECUTION]: There is no evidence of remote script downloads or dynamic code execution. The skill refers to an internal
writing-planstool for subsequent steps. - [COMMAND_EXECUTION]: The skill does not invoke arbitrary shell commands or attempt to gain elevated privileges.
Audit Metadata