The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes project tasks using detected runners (mise, just, task, npm/pnpm). All execution is channeled through a single, controlled entry point (scripts/task.sh) and uses exec to forward arguments correctly to the detected runner.\n- [SAFE]: The skill explicitly includes a human-in-the-loop requirement for creating or modifying project configuration files (e.g., mise.toml, justfile, package.json). It offers snippets to the user rather than writing them silently.\n- [SAFE]: Arguments are handled carefully; the script uses literal matching for task resolution and forwards remaining arguments verbatim to the task runner, reducing command injection risks from user input.\n- [SAFE]: The skill implements error handling for untrusted configurations (specifically for mise), directing users to verify trust status (mise trust) rather than bypassing safety mechanisms.

tasks