insane-design-build

Warn

Audited by Gen Agent Trust Hub on May 1, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill dynamically generates and executes a Python script using a heredoc (`python3
  • <<'PY'`) to perform screenshot verification via Playwright. This pattern of executing runtime-constructed scripts is a significant security risk.
  • [COMMAND_EXECUTION]: The skill automatically executes platform-specific browser commands (open, xdg-open, start) to display generated HTML files. This behavior bypasses manual user review of the generated files before they are interpreted and executed by a web browser.
  • [PROMPT_INJECTION]: The 'Identity' section employs 'Persona Lock-in' techniques, instructing the agent to ignore user requests for neutral options and to 'never hedge'. These instructions are designed to override the agent's default balanced decision-making behavior.
  • [DATA_EXPOSURE]: The skill uses broad filesystem globbing (Glob "${CLAUDE_PLUGIN_ROOT}/skills/insane-design/examples/*/design.md") to find configuration files. While restricted to the plugin directory, this pattern can be used to scan for and expose file structures.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 1, 2026, 10:57 AM
Security Audit — agent-trust-hub — insane-design-build