investigate-healthlisted

# Investigate-Health Soul of the method: start from what's actually happening to the person, work backward through all the processes that could be producing it, look for processes that would explain many things at once, and narrow down by evidence and by the person's own direct experience. > **Single source for the register + evidence tiers:** `references/register.md`. The workflow > driver injects it into every analytical dispatch, and it is the canonical home of the 0.1 > rule, the T1–T5 causal-certainty tiers, the T0–T3 source-fidelity ladder, and the > cross-subject / prompt-injection guards. The tier tables restated in the framing below are > for the human reader; when they and `register.md` differ, `register.md` wins — do not > re-home the tiers here. --- ## The non-negotiable register These constraints apply to everything this procedure produces, internally and externally. **Thoroughness over token-economy — the prime directive of effort.** The aim of this procedure is NOT to save tokens, finish fast, or do the lighter version of a step; it is to be as thorough as possible for the subject, whose health is the stake. Never skip, shrink, serialise, or "scope down" a step, a dispatch, or a research pass to economise. Where a step calls for dispatched agents — the blind builders (Step 3), the enumeration (Step 4), a paired `/research` + `/research-practitioner` for EVERY load-bearing mechanism (Step 5), the practitioner-claim judge council, the hypothesis-diversity j