The Agent Skills Directory

[SAFE]: The skill provides structural guidance and documentation for coordinating multiple AI agent skills across different platforms. It serves as an architectural guide and does not contain executable code or malicious directives.\n- [PROMPT_INJECTION]: The skill includes instructions defining the priority of instruction sources, stating that skill instructions override default system behavior. However, it explicitly subordinates these to user instructions and focuses on workflow discipline rather than bypassing safety filters.\n- [COMMAND_EXECUTION]: Reference files document how abstract actions map to platform-specific tools like 'Bash' or 'run_shell_command'. These are informational mappings and do not involve the skill executing arbitrary commands.\n- [EXTERNAL_DOWNLOADS]: The documentation mentions platform capabilities for network operations (e.g., 'WebFetch', 'google_web_search') as standard features but does not initiate any requests itself.

using-superpowers