The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a provided Node.js script (assets/build-status-data.cjs) to manage the deployment of web assets and start a local HTTP server for its decision-making UI. The script includes significant security measures, such as containment checks using fs.realpathSync to ensure file operations do not escape the project root and checks to prevent the ingestion of symlinks or hard links.
[DYNAMIC_EXECUTION]: The skill core functionality involves the execution of its own Node.js assets to perform file system migrations (clearing obsolete files and moving artifacts) and serving local content. The HTTP server is strictly bound to the loopback address (127.0.0.1) and only supports read-only GET and HEAD methods.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-generated or prior-AI-generated Game Design Documents (GDD) and Phase plans from the project's Docs/ directory to facilitate resuming work across sessions. This creates a surface for potential instruction injection. To mitigate this risk, the skill mandates a 'hard gate' requiring the agent to re-read its authoritative management rules (references/build-management.md) before updating any task states or project records.

msw-planning