using-superpowers

To enforce skill invocation before agent behavior, you need pre-action workflow guardrails that check for applicable skills. This requires invoking the Skill tool whenever there is even a 1% chance of relevance, ensuring domain workflows run before any response is generated.

Your agent skips clarifying questions due to red-flag rationalization, where thoughts like 'this is too small' bypass tool routing. Applying prompting discipline stops the agent and forces a skill check whenever it considers skipping a potentially relevant workflow.

The best way to prioritize process skills over implementation skills is to apply skill selection priority guidance. This orders process-oriented skills like brainstorming or debugging before design or execution skills when multiple workflows could apply to the conversation.

You need pre-action workflow guardrails at all starting points of a conversation, including early workflow decisions and clarifying questions. This prompting discipline ensures you never proceed without checking whether a specialized skill applies to the user's request.

Yes, you can use prompting discipline to stop agents from replying prematurely by enforcing a pre-action skill invocation rule. This rule intercepts the conversation before any response, requiring the agent to run a skill check and follow triggered instructions.

Without workflow guardrails, agents often fail to self-select skills due to rationalizing marginal relevance. Even when a request feels small or context seems insufficient, an agent might bypass tool routing, missing specialized domain workflows that should have been triggered.