Provenance-first review
Review output should preserve source references, artifact context, and evidence state so examiners can return to source-native material.
Defensibility
Preserve provenance, validation state, and unresolved gaps.
TraceLodger is designed around review-ready evidence handling. It distinguishes direct artifacts, bounded inference, warnings, unsupported states, contradictions, and uncalibrated parser output.
DirectInferredUnsupportedWarningContradictionUncalibrated
Endpoint artifact
Retained source reference
AI-session trace
Bounded continuity cue
Parser output
Awaiting validation gate
Export conflict
Reviewer attention needed
Validation gates
Parser-backed claims stay marked until validation gates support review-ready use. Uncalibrated states remain explicit.
False-positive controls
Inferences are separated from direct artifacts, and warning states signal where a reviewer should avoid stronger conclusions.
Parser regression
Parser behavior needs repeatable checks so format changes, source drift, and unsupported records do not become quiet claims.
Review-ready export bundles
Exports should carry source references, evidence states, unresolved gaps, and clear separation between local evidence and enrichment.
Contradiction handling
Conflicting records are surfaced for examiner attention instead of being flattened into a single timeline story.
Boundary conditions
What the software will still leave unresolved
TraceLodger should make unresolved areas easier to review, not less visible. Gaps stay part of the record so reviewers know where source material remains primary.
Warning
Intent behind user behavior.
Warning
Activity outside retained endpoints, browsers, mobile sources, logs, or exports.
Warning
Continuity when required artifacts are missing, damaged, or contradicted.
Warning
Claims from parsers that have not passed validation gates.
Review posture
Use evidence states as product language and reviewer guidance.
TraceLodger should remain precise about source support, bounded interpretation, contradictions, and validation status before a reviewer treats output as review-ready.