Review of an existing LLM instructions library.

I have developed an AI coaching system with a mature coaching methodology and seven core behavioural instruction documents plus a rules box.

I am looking for an experienced LLM Systems Architect, AI Behaviour Designer, or Senior Prompt Engineer to review the instruction library and provide recommendations.

This is not a software development project and not a request to build an AI coach from scratch.

The objective is to review the existing behavioural architecture and identify:

* conflicting instructions;
* redundant instructions;
* instructions that are too weak, abstract, or ambiguous;
* instructions that are unlikely to produce the intended behaviour;
* behaviours that cannot be reliably achieved through prompting alone;
* opportunities to simplify or restructure the architecture.

The main behavioural challenge is that the AI still tends to:

* accept the user's framing too quickly;
* reason inside the user's subjective account of the situation;
* stop exploring too early;
* miss important information that has not yet entered the conversation.

Deliverables:

* Review of the seven documents and rules box.
* Written recommendations and suggested changes.
* Explanation of why specific behaviours are not emerging consistently.