name: research-routes description: Compare multiple research routes with explicit Claims, Evidence, unknowns, failure regimes, and reversible probes. Use when a person needs route-neutral exploration before implementation, experiment authorization, HumanDecision, or canonical-state promotion. metadata: version: "1.0.0"

Research Routes

Keep a route-neutral terrain map separate from decisions and execution.

1. Inventory candidate routes and source-backed evidence.
2. Keep observations, interpretations, Claims, Evidence, contradictions, and unknowns distinct.
3. Compare assumptions, mature-method comparators, costs, failure regimes, and falsifiable probes.
4. Preserve every evidence polarity, including support, counter, null, boundary, conflict, failure, context, mixed, and unknown.
5. Choose only a reversible next probe when the user authorizes exploration.
6. Record FailureRecord, RewindProposal, Lesson, ReopenedProblem, and canonical downgrade as append-only research records; never erase the failed path.
7. State that a terrain map is not a route choice, experiment authorization, implementation, HumanDecision, or canonical-state promotion.

The Pack runtime lives under runtime/research_routes_runtime/ and compiles research-pack/v1. It uses only generic host, artifact, snapshot, decision, and promotion boundaries supplied by All in Luna; it does not add research state to Core. A route authorization must name a confirmed HumanDecision, and canonical promotion needs its own explicit decision scope.

When a route is authorized for product work, hand off its bounded decision and evidence package to $allinluna; do not silently turn research notes into implementation tasks.