﻿LOGIC OR NOT P075 — MONITOR RUNTIME
Version: v0.1.1-CANDIDATE
Status: CONFIGURED_NOT_ACTIVE / TEST-READY / UNPROVEN
Governed object: tracked claim/evidence state versus newly observed evidence and source lifecycle changes.


PRIMARY TEMPORAL FRAME
Trigger/preflight -> verify claim/source/state access -> load claim/evidence baseline -> acquire current sources -> normalize source/evidence identity and lifecycle -> compare -> materiality -> notify or remain silent -> transactional evidence/runtime writeback -> continue/stop.


RUN RULES
1. Read installed MONITOR_CONFIG and MONITOR_STATE. Missing trigger, source registry, evidence-state binding, delivery route or expected state revision is NOT_CONFIGURED, SOURCE_FAILURE or BLOCKED; never NO_CHANGE.
2. Track stable claim_id plus claim_version. A material change to claim wording/scope is CLAIM_DEFINITION_CHANGE and must not be interpreted as evidence movement against an unchanged claim.
3. Preserve strongest source and evidence identity available: source_id/publication_id, source_revision/version, evidence_unit_id and stable locator. A correction, update, withdrawal or retraction of an existing source is a lifecycle event, not automatically new independent evidence.
4. Separate source, evidence unit, evidence-to-claim relation and claim state. Record relation direction, support status, evidence state, provenance and observed_at. Do not collapse source existence into support.
5. Detect duplicate/syndicated evidence and independence. Re-publication or syndication of the same underlying evidence cannot increase independent support count.
6. Version the source registry and freshness policy. SOURCE_REGISTRY_CHANGE is distinct from EVIDENCE_CHANGE; newly configured sources do not retroactively prove a claim.
7. Classify claim-level changes before materiality: NEW_SUPPORT, WEAKENED_SUPPORT, CONFLICT, SUPERSEDED, CORRECTION, RETRACTION, CLAIM_DEFINITION_CHANGE, SOURCE_REGISTRY_CHANGE, SOURCE_UNAVAILABLE or NO_MATERIAL_CHANGE.
8. Maintain evidence_baseline_current, evidence_baseline_history and evidence_baseline_candidate separately. Candidate state changes require the configured owner/adoption authority.
9. Commit a run ledger entry atomically: run_id, claim set/version, source_registry_version, freshness coverage, source/evidence revisions, baseline version read, expected state revision, delta classes, materiality, notification result, owner/runtime writeback result and resulting state revision.
10. Build change_fingerprint from stable claim/evidence/lifecycle identities and normalized material delta. Suppress only the same already-notified fingerprint.
11. Runtime state is not the owner research/evidence registry. Direct evidence-state writeback requires an authorized owner route; otherwise emit exact EVIDENCE_STATE_DELTA_TO_APPLY with expected owner revision.
12. Silence is valid only after a successful freshness/coverage check across the configured source scope and no material delta. Partial source failure or unresolved claim comparability forbids a no-change conclusion.
13. Never upgrade confidence from unavailable or duplicated evidence. Preserve UNKNOWN and contested states.
14. Consequential external action remains outside this Monitor unless separately governed.


OUTPUT ON MATERIAL CHANGE
Tracked claim and version; evidence/source identity and lifecycle; prior versus current evidence relation/state; delta class; independence/provenance; material impact; confidence/unknowns/conflicts; review/action if supported; fingerprint; writeback result.


OUTPUT ON FAILURE
Name affected claim/source/coverage, last successful observation if known, and what cannot be concluded. Never translate failure into no change.
