Evidence Normalization

Operation

Normalize the supplied evidence without changing the product or selecting an implementation.

Goal

Convert heterogeneous repository findings, logs, traces, tickets, and audit outputs into a deduplicated evidence set suitable for synthesis without changing priority or inventing missing facts.

Normalize

Accept logs, audit prose, tickets, traces, plans, or model outputs. Preserve provenance and exact uncertainty. Extract only evidence-backed owner, invariant, impact, acceptance, validation, scope, and rollout data. Do not upgrade hypotheses to facts, merge independent owners, or choose downstream implementation work.

Return the narrowest applicable artifact: intake summary, audit result, task definition, implementation summary, or check evidence. When source material conflicts, keep both claims and mark the missing discriminator.

• Preserve exact path:line[-line], symbol, command, trace, timestamp, environment, and source references when available.
• Mark each statement as direct evidence, inference, assumption, or unavailable fact.
• Merge only identical observations; keep contradictory evidence side by side for Synthesis.
• Remove prose that does not change owner, invariant, impact, acceptance, validation, dependency, or exclusion.

Keep each observation attached to its source revision/environment and the claim it actually supports. A successful local fixture does not cancel a production failure from another environment. The handoff to Synthesis contains deduplicated observations, unresolved contradictions and missing discriminators; it does not label hypotheses as implementation-ready findings.