Stage 6 · detail
Playbook catalog
The hypothesis engine walks 17 playbooks across 8 precedence groups. For each cluster it evaluates every playbook against the evidence predicates and emits ranked candidates — the highest-precedence match becomes the primary hypothesis the LLM narrates. This page explains exactly what fires each one and what it is allowed to recommend.
How a playbook is evaluated
Each card lists the predicate conditions a cluster's evidence must satisfy, plus the recommendation templates the playbook unlocks.
required all must be satisfied
any-of at least one
stronger raises confidence & can reach confirmed_issue
exclusion any match → playbook skipped
soft-contrary lowers the alternatives factor
template allowed recommendation
Precedence groups
Groups compete in order. A later group still contributes alternates (rank ≥ 2) but won't supply the primary if an earlier group already matched. Fallback playbooks only emit when nothing else claimed the cluster.
flowchart LR
G1["1 · Measurement"] --> G2["2 · Technical"] --> G3["3 · Interaction"] --> G4["4 · Discovery"]
G4 --> G5["5 · Merchandising"] --> G6["6 · Trust"] --> G7["7 · Scope"] --> G8["8 · Fallback"]