The Agent Skills Directory

[NO_CODE]: The skill consists entirely of Markdown guidelines (SKILL.md, LOGIC.md, UI.md) and configuration metadata (.provenance, agents/openai.yaml). There are no scripts, executables, or automated commands that run on the host system.\n- [SAFE]: The instructions promote secure prototyping practices, such as using in-memory state or local scratch databases rather than production persistence, and gating UI debug tools behind environment checks (process.env.NODE_ENV !== 'production').\n- [REMOTE_CODE_EXECUTION]: While the skill mentions running commands like pnpm, python, or bun, these are instructions for the human user to execute their own local project tasks; the skill does not perform these actions automatically or download external code.\n- [COMMAND_EXECUTION]: The skill documents the platform-specific !command syntax but does not use it to execute any hidden or dangerous shell commands.

prototype