research-knowledge-core

name	research-knowledge-core
description	Shared reference for the research-knowledge cluster: the evidence ladder (lightest lane first → escalate on synthesis need), the four provenance tiers every claim carries, source/freshness conventions, and the question-type → spoke matrix. USE WHEN running any research, literature, IP, genomic, codebase, or docs-lookup task — the discipline every spoke in this cluster shares.
cluster	research-knowledge
version	1.0.0

Research & Knowledge Core

Shared model for the research-knowledge cluster. Every research, literature, database, codebase, and reference spoke depends on these two interlocking ideas — keep them consistent here so no spoke contradicts another.

1. The evidence ladder (this cluster's defining discipline)

Answer with the lightest lane that actually answers the question, and escalate only when a heavier one is justified. Climbing a rung costs time, tokens, and quota — earn it.

local context / docs / ck memory   ← cheapest, no network
        │  (not present or stale?)
        ▼
fast discovery (exa-search, single DB query)
        │  (one source not enough? need synthesis?)
        ▼
multi-source synthesis (deep-research, systematic literature review)
        │  (claim needs judging?)
        ▼
evaluation / verification (scholar-evaluation, citation checks)

Rule: never start a heavyweight pass when a lookup answers it; never stop at one source when the claim is load-bearing and contested. research-ops is the operator that picks the starting rung — route ambiguous "research this" asks there first.