Exact Evidence Locators and Provisional Claims in a Local Investigation Workbench

An investigation tool can make a claim look well sourced while obscuring whether its citation points to preserved content, an unvisited reference or an automated suggestion. This paper describes Trace, a local workbench that preserves supplied bytes and hashes, validates selected evidence locators, distinguishes user classifications from provisional lexical matches, and exports a report with a receipt register. The study is a retrospective design and source-inspection report, not a measured analyst-performance evaluation. The inspected implementation uses case-scoped SQLite storage, exact stored-text offsets and bounded image regions. Public discovery is opt-in; returned references remain distinct from separately captured pages. A preserved fictional demonstration contains six receipts and shows how claims and timeline entries link to different evidence types. Source tests specify rejection of invented excerpts, out-of-bounds regions and cross-case links, but no new test execution or accuracy benchmark is claimed here. The design makes some evidential dependencies inspectable. It does not authenticate the original source, determine whether an allegation is true, supply a tamper-proof chain of custody or demonstrate access to justice. The paper identifies the remaining evaluation requirements for contestability, privacy, correction and practical user control. Preparation of this report was assisted by Codex. The research programme is directed toward human agency, financial inclusion, and evidence-based accountability. Intended social benefits are distinguished from measured technical results in the manuscript. The manuscript and newly authored analysis are offered under CC BY 4.0. Any supplied source files retain the license terms stated in the supplement; the record license does not replace third-party source licenses.