skills/0x2e/superpowers/brainstorming/Gen Agent Trust Hub

brainstorming

Warn

Audited by Gen Agent Trust Hub on Mar 26, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Executes custom Bash scripts (start-server.sh, stop-server.sh) and a Node.js script (server.cjs) to manage the 'Visual Companion' server.
  • [COMMAND_EXECUTION]: Uses system-level commands like ps and kill to perform process monitoring and lifecycle management of the background server.
  • [COMMAND_EXECUTION]: Instructs the agent to use platform-specific execution flags (e.g., run_in_background: true) to ensure the server process persists across multiple conversation turns.
  • [DATA_EXFILTRATION]: Spawns a local web server that listens on a random high port. While it binds to the loopback interface (127.0.0.1) by default, documentation suggests binding to 0.0.0.0 for remote or containerized environments, which could expose the server to the network.
  • [DATA_EXFILTRATION]: The server reads and serves files from the local filesystem (specifically a session-specific directory) to allow the user to view agent-generated HTML mockups in a browser.
  • [PROMPT_INJECTION]: Utilizes strong instructional framing, including <HARD-GATE> blocks and absolute directives (MUST, Do NOT), to override default agent behavior and enforce a strict design-before-implementation policy.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 26, 2026, 01:29 PM
Security Audit — agent-trust-hub — brainstorming