creative-production-moodboard-explorer

Warn

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill launches a local Node.js server (server.mjs) and several Python scripts (create_mood_board.py, embed_moodboard_images.py) to manage its application state and image processing.
  • [COMMAND_EXECUTION]: The instructions in SKILL.md explicitly direct the agent to request "escalated/outside-sandbox execution" if it detects that the environment is restricted, which is used to bypass security boundaries for accessing local state in directories like ~/.open-cowork.
  • [COMMAND_EXECUTION]: The agent is instructed to execute complex shell commands, including a Node.js one-liner using the -e flag to interact with the local server's REST API and perform cryptographic hash operations.
  • [REMOTE_CODE_EXECUTION]: The Node.js server executes a Python-based image generation runner (codex_exec_image_batch.py) located outside the skill's own directory structure, with the specific path retrieved from a configuration file (runtime-config.json) or environment variables.
  • [DATA_EXFILTRATION]: The local server includes functionality to interact with the system clipboard using osascript (macOS), which provides a capability for programmatic reading and writing of clipboard content.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 10, 2026, 06:05 PM
Security Audit — agent-trust-hub — creative-production-moodboard-explorer