The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes shell scripts (start-server.sh, stop-server.sh) and a Node.js script (server.cjs) to facilitate a local brainstorming companion server. These scripts manage local process execution, including backgrounding with nohup and PID tracking for cleanup.
[EXTERNAL_DOWNLOADS]: The documentation refers to several external design and motion reference websites (e.g., MotionSites AI, 21st.dev, React Bits, GSAP) as research sources. The scripts/README.md file provides a command-line snippet for manually re-syncing scripts from a GitHub repository (obra/superpowers), intended for developer maintenance.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data such as "Product Briefs" and project source code to generate a "Build Contract". It also reads interaction events from the local companion server's state directory. The instructions emphasize verifying intent and using native input interfaces to mitigate potential injection risks from untrusted content.
Ingestion points: Product Brief, project sources of truth, and $STATE_DIR/events.
Boundary markers: Present (instructions to verify intent, resolve only what's needed).
Capability inventory: Shell script execution, file writing, and Node.js server operations.
Sanitization: JSON parsing for events and escapeHtmlText in branding elements.
[DYNAMIC_EXECUTION]: The companion server (server.cjs) uses Node.js child_process modules to open the user's web browser. It utilizes execFile for platform-default launchers and exec for an optional operator-provided launcher command specified via the BRAINSTORM_OPEN_CMD environment variable.

deep-design