Bohan Jia

Classify the action type for a topic loop step using semantic reasoning instead of keyword matching. Load when the action queue is being materialized.

Use when an AITP topic needs a clear statement of what is currently believed, what is only provisional, what is blocked, and what still needs work.

Use when AITP needs to write or review a derivation so the result reads like a serious theoretical-physics argument rather than polished prose or a loose sketch.

Use when one bounded AITP research round should be written as a real L3-L4 scientific cycle with a question, plan, check, result, and next-step decision.

Determine whether a topic should use the full or light runtime load profile, using semantic reasoning instead of regex matching.

Use when an AITP topic needs to be written as a real theoretical-physics problem with motivation, literature position, scope, and non-goals instead of a generic runtime question.

Classify the research mode of a topic using semantic reasoning instead of keyword matching. Load before session-start or resume to ensure research mode is recorded.

Select the runtime mode and active submode for a topic using semantic reasoning instead of keyword matching.

Use after AITP routing has claimed the task; continue theory work through the runtime bundle instead of ad hoc browsing or free-form synthesis.

Use when an AITP topic should be compiled into a physicist-readable report with a clear scientific main text and appendix-style audit material.

Use when a request might be theoretical-physics research, topic continuation, idea steering, paper learning, derivation work, or validation planning; enter AITP before any substantial response.