using-superpowers

To enforce LLM agent tool use, you need a workflow protocol that mandates skill invocation before any text generation. This ensures the agent always checks for applicable tools, preventing missed automation opportunities and improving accuracy.

AI agents skip available skills due to rationalizations that a skill is unnecessary. Enforcing a strict '1% chance' rule overrides this behavior, requiring the agent to invoke any potentially relevant process or implementation skill before answering.

Prioritizing process skills over implementation skills requires defining a strict invocation order within your agent workflow. This ensures foundational process rules execute before specific implementation tasks, maintaining protocol adherence across conversational interactions.

To stop an LLM from rationalizing skipped tool use, apply clear skill usage guidelines that identify these rationalizations as red flags. Mandatory invocation rules force the agent to attempt tool use whenever there is even slight applicability.

Mandatory skill invocation is needed for AI conversational agents whenever a user query could potentially be assisted by a tool. Applying this protocol to all task-oriented interactions ensures consistent automation and prevents the agent from relying solely on internal knowledge.