The Agent Skills Directory

[COMMAND_EXECUTION]: The skill manages a local Node.js web server via scripts/start-server.sh and scripts/stop-server.sh. This server, implemented in scripts/server.cjs, uses native Node.js modules (http, fs, path, crypto) to provide a visual interface for mockups.
[COMMAND_EXECUTION]: The server binds to the local interface (127.0.0.1) by default and employs path.basename to sanitize requests for files in the content directory, preventing directory traversal attacks.
[COMMAND_EXECUTION]: Security-conscious lifecycle management is present; the server monitors the agent's process ID (OWNER_PID) and includes an idle timeout (30 minutes) to prevent persistent background processes after the session ends.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests untrusted data from project files, documentation, and git commits (Ingestion points: SKILL.md). It lacks specific boundary markers or sanitization for this content. However, its capability tier is limited to design documentation and planning (Capability inventory: writing to docs/, server execution, and writing-plans invocation), resulting in a low-risk profile.

brainstorming