Scope and Non-Goals

bijux-proteomics-intelligence owns inspectable decision support over a fixed candidate set and evidence revision. It explains why an option progressed, held, lost, or was refused under a named policy. It does not manufacture evidence or authorize execution.

In Scope

• candidate eligibility, hard filters, ranking, ties, and exclusions;
• evidence-readiness posture and decision-specific insufficiency;
• weights, thresholds, score orientation, missingness, and policy identity;
• confidence, calibration context, sensitivity, counterfactuals, and regret;
• contradiction, falsifier, skeptical-review, and human-escalation records;
• progression, hold, redesign, synthesis, and scale-up judgments;
• advisory envelopes, explicit promotion to enforced policy, and refusal;
• planned-versus-observed review and attributable policy-learning proposals.

Explicit Non-Goals

Not owned here	Responsible boundary
parsing, quantification, FDR, scientific acceptance, and benchmark truth	Core
process execution, provider selection, state, replay, and artifacts	Runtime
source custody, identity resolution, claims, and contradiction evidence	Knowledge
assay design, readiness, resource commitment, handoff, and observations	Lab
identity, serialization, provenance primitives, and schema migration	Foundation
release approval and repository policy	maintainer and release governance

Intelligence also does not own autonomous progression. A high score, apparent consensus, or calibrated confidence value remains advisory until a separate authority promotes the decision under a named policy.

flowchart LR
    evidence["versioned evidence"] --> candidates["eligible candidates"]
    candidates --> policy["declared decision policy"]
    policy --> challenge["sensitivity and skeptical challenge"]
    challenge --> posture{"decision posture"}
    posture --> advisory["advisory recommendation"]
    posture --> hold["hold, redesign, or refusal"]
    advisory -. "separate authority" .-> enforced["enforced policy"]

Ownership test

A behavior belongs here when changing it can change which defensible action is preferred while the scientific result and evidence records remain fixed. A behavior belongs elsewhere when it changes the observations, execution history, source meaning, or physical feasibility on which that judgment depends.

Every public decision must retain the candidate universe, exclusions, evidence revision, policy, challenge coverage, uncertainty, alternatives, refusal boundary, authority state, and unresolved questions. Without that packet, the output is a ranking result—not reviewable decision support.