using-superpowers

To stop an AI agent from bypassing tools, enforce a strict skill invocation protocol requiring agents to check and invoke applicable skills before generating any response. This addresses the problem of agents making assumptions or ignoring workflows.

An AI agent ignores available skills when it rationalizes that direct answers are sufficient. Enforcing a mandatory invocation workflow with a 1% chance rule for skill applicability ensures the agent checks tools even if there is only a slight possibility they apply.

The correct skill invocation workflow requires an AI agent to always invoke relevant skills before performing any action or responding. It prioritizes process and implementation skills to ensure proper tool usage and adherence to established protocols.

To prioritize process skills over implementation skills, apply defined skill prioritization guidelines within the invocation protocol. This ensures the AI agent addresses workflow processes first before executing specific implementation tasks.

A limitation of enforcing mandatory tool usage is overcoming the AI agent's common rationalizations for avoiding skills. The protocol addresses this by providing clear workflow guidance and flagging these rationalizations to ensure compliance.