using-superpowers

To enforce AI safety guardrails before every conversation reply, you need a policy that mandates Skill tool invocation at the start of every interaction. This ensures appropriate skill guidance is loaded and applied before generating any response.

The best way to prevent unchecked or unsafe AI responses is to enforce pre-response Skill tool invocation across all conversations. This policy loads relevant skill content before replying, ensuring correct guidance and safety guardrails are applied.

To load the correct skill guidance during clarifying interactions, invoke the Skill tool whenever a skill may apply. This ensures the applicable skill content is loaded and applied before the system responds to the user.

You need to invoke a skill tool for context-aware planning at the start of the interaction. The system first loads the applicable skill and then generates a response based on its specific instructions and guardrails.

Yes, you can apply skill invocation guardrails to prevent prompt injection attacks. By requiring the Skill tool to load and apply correct skill content before any response, the system ensures proper guidance and safety throughout the conversation.

This skill invocation policy applies to every AI response, not just the first reply. It enforces Skill tool invocation at the start of all conversations and during clarifying interactions whenever a skill may apply.