using-superpowers

To ensure an AI agent checks for applicable skills before responding, enforce a mandatory skills-first workflow that requires invoking the Skill tool before any reply or action, even for clarifying questions. This guarantees instruction compliance across every conversation turn.

An agent skips relevant instructions when multiple workflow skills apply because it lacks skill selection discipline. Enforcing a very low applicability threshold of 1% for invoking the Skill tool ensures the correct operational instructions load before answering.

The best way to enforce prompt discipline in an agent workflow is to require a pre-response skill check. This process mandates invoking the Skill tool instead of a file Read tool, ensuring rigid or flexible skill handling and priority rules are followed before any response.

To trigger skill invocation for Protheus or ADVPL engineering tasks, instruct the agent to immediately use the Skill tool upon receiving the user message. This verifies whether a specialized engineering workflow skill applies before generating an answer.

You should use process skills versus implementation skills based on defined priority rules within the agent workflow. The skills-first mechanism provides specific guidance on choosing between process and implementation instructions to maintain prompt discipline.

Yes, the skills-first approach applies to clarifying questions during a conversation. The enforcement mechanism operates on every conversation turn, requiring the agent to check for applicable skills before asking for clarification or taking any action.