using-superpowers

To enforce tool use in an LLM agent, you need a mandatory skill invocation protocol that requires the agent to check and execute relevant tools before generating any output or taking action. This ensures consistent tool utilization across all conversational turns.

An AI agent skips tool invocation when it lacks a strict prioritization logic for skill execution. By enforcing mandatory invocation even with a low probability of applicability, the agent is required to check relevant process and implementation skills before attempting a direct response.

The correct prioritization logic for agent protocol execution dictates that process skills are invoked first, followed by implementation skills. This ordering ensures that overarching process rules are applied before specific implementation tools are executed.

Yes, you can require an LLM to always check for relevant skills by applying a strict skill invocation protocol across all conversational turns and task executions. This guarantees the agent evaluates tool applicability before responding.

Yes, mandatory skill invocation works even with a low probability of applicability by enforcing a strict protocol that requires the agent to invoke potentially relevant skills before generating a response. This prevents the agent from skipping potentially useful tools.