brainstorming
Pass
Audited by Gen Agent Trust Hub on Jul 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Starts a local Node.js server using
scripts/start-server.shto provide a visual brainstorming companion. The server manages mockups and user interactions across conversation turns.\n- [EXTERNAL_DOWNLOADS]: Fetches a branding logo fromprimeradiant.comto display in the companion browser interface.\n- [DATA_EXFILTRATION]: The companion server binds to a local port (default 127.0.0.1) to serve visual content. Security is maintained through a session-specific token requirement and origin validation to prevent unauthorized access from other local processes or sites.\n- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting untrusted data from project files, documentation, and commit history. \n - Ingestion points: Project files and git history read during the 'Explore project context' step in
SKILL.md.\n - Boundary markers: None explicitly defined in the instructions to separate project data from agent instructions.\n
- Capability inventory: The skill can write files to the project directory, execute shell scripts to manage the companion server, and invoke the
writing-plansskill.\n - Sanitization: No specific sanitization or filtering of the ingested project data is performed before it is processed by the agent.
Audit Metadata