The Agent Skills Directory

[COMMAND_EXECUTION]: The skill manages a multi-phase generation workflow using shell commands.\n
Evidence: Uses mv and rm -rf for atomic file operations and cleanup of temporary staging directories.\n
Evidence: Executes chmod +x on the generated run-goals.sh script to enable its execution.\n- [REMOTE_CODE_EXECUTION]: The skill generates scripts and configurations that execute external commands.\n
Evidence: Produces a run-goals.sh bash script that executes codex exec to drive the autonomous implementation of goals.\n
Evidence: The generated script defaults to a high-privilege sandbox mode (danger-full-access) to accommodate Docker and network access.\n- [EXTERNAL_DOWNLOADS]: The skill references and installs tools from external sources.\n
Evidence: Instructs users to install @playwright/cli and @openai/codex via global NPM installation.\n
Evidence: References WordPress plugin ZIP files (e.g., WooCommerce) for installation in the wp-env environment.\n- [PROMPT_INJECTION]: The skill processes external requirements files to generate project artifacts.\n
Ingestion points: requirements.md, notes/ directory.\n
Boundary markers: Employs a specific 'sanitizer' protocol to detect and block platform-specific tokens and destructive commands.\n
Capability inventory: File system writes, shell command generation, and script execution via run-goals.sh.\n
Sanitization: Filters for Claude Code references, destructive shell tokens (wp plugin uninstall, DROP TABLE), and absolute host paths.

wp-requirements-to-goals