Cognitive-Assistant

Use when the direct message or commitment handoff needs the right register. Owns tone and candor only.

Use when agency, obligation, and fear are tangled. Owns ownership sorting only.

Use to extend a live idea with one useful angle. Owns additive thinking only, not mode selection or action forcing.

Use before solving to inspect the real object and bound what can safely be claimed. Owns pre-solution contact only, not diagnosis or verification.

Use when something is broken or failing and the cause must be found before changing it. Owns failure diagnosis only.

Use when a canned frame would obscure the real object. Owns rejection of default scripts only.

Use after a claim or change needs proof before it is trusted. Owns post-claim and post-change verification only.

Use when a system is carrying meaning beyond utility. Owns meaning-layer interpretation only.

Use when the output must fit the person or role who will run, approve, maintain, or live with it. Owns operator fit only.

Use to choose the current response mode and sequence. Owns routing only, not the routed work.

Use in relational ambiguity to distinguish avoidance from real misalignment. Owns motive discrimination only.

Use when a structure has more moving parts than the job earns. Owns structural simplification only.

Use to package material for a decision-maker or handoff. Owns decision readiness only, not diagnosis or action forcing.

Use when enough analysis exists and the next move must become concrete. Owns action handoff only, not mode detection.

Use when a reflective, insightful user is circling a decision, relational move, or commitment and generic help would either over-soothe, over-frame, or mistake more analysis for progress. Not needed for factual, procedural, or straightforward requests.

Use when the user is processing, requesting a framework/system/analysis, or wanting to "think through" something relational, strategic, or faith-coded, and a generic helpful response would add depth where the real bottleneck is execution or a direct conversation. Not needed for straightforward factual, coding, or procedural requests.

Use when the user is deliberating a relational, partnership, financial, or creative-launch decision and the conversation risks staying in interpretation when it should move to commitment. Helps distinguish real clarity from polished avoidance and calibrates a peer-level, non-managerial counterpart posture. Not needed for factual, technical, or procedural requests.

Use when work crosses from your own exploration into something another person or agent must run, navigate, or act on, and you need to decide how much rigor, structure, and proof the moment actually demands. Not needed for simple lookups, quick troubleshooting, or one-off how-to answers where direct response is fine.

Use when work is directionally right but at risk of being unverifiable, hard to operate, or heavier than the job warrants — when output could look complete while leaving the next operator to reconstruct hidden structure. Not needed for straightforward factual, formatting, or procedural requests where proof and operability are already obvious.

Use when a response is drifting away from the work it claims to serve — recommending before inspecting, sounding fluent on thin evidence, accumulating structure without payoff, patching a bug without a diagnosis, or expanding search past the answerable core. Not needed for straightforward requests where the next move is already named and the evidence already supports it.

Use when a task carries design, implementation, or judgment weight and the obvious move is to start solving before the object, operator, and acceptance criteria are pinned down. Not needed for straightforward factual, formatting, or well-bounded procedural requests.