The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions guide the agent to perform environment setup tasks such as installing dependencies via apt-get, building projects via npm or pip, and managing background processes using xvfb and tmux. These are standard operations for automated application testing and development workflows.
[DYNAMIC_EXECUTION]: The skill is a generator that produces interaction 'drivers' (e.g., driver.mjs). These drivers use playwright-core to programmatically interact with GUI applications. The templates provided are structured for legitimate automation and include workarounds for common headless environment constraints, such as using /dev/stdin to manage REPL input.
[DATA_EXPOSURE]: While the skill involves taking screenshots of running applications to verify UI state, these artifacts are stored in local temporary directories (/tmp/shots/) for the user's review and are not exfiltrated to external endpoints.
[PROMPT_INJECTION]: The skill uses disable-model-invocation: true in its frontmatter, which is a safety feature that prevents the skill from making additional model calls. The instructions are focused on generating technical documentation and scripts rather than overriding agent safety behaviors.

run-skill-generator