# DN Evidence Retest Planner v0.1

Initial release: 5 November 2026. Six illustrative claim types and fourteen invented change scenarios. No model calls, provider trials, paid requests, safety audit or evidence-authenticity verification. This is an editorial dependency map, not a standard, certification expiry rule, exhaustive risk model or deployment gate.

The planner flags claim scopes for review when a known changed dependency intersects the published map. A model change flags accepted-output quality, completion time and outcome cost in this example. Pricing alone flags cost. A cosmetic label-only change flags nothing under this map's assumptions. That is not evidence of safety or current performance. Unknown changes produce unresolved scope across all six claims, not a finding that six tests failed. Unknown dominates a mixed known/unknown input.

Historical evidence remains a record of its stated configuration and scope. No automatic age threshold or validity duration is calculated. Claims not flagged are not cleared; the map can omit relevant dependencies. A qualified owner must check completeness, test sufficiency, operating context and the consequences of relying on a claim.

The source accepts one to sixteen unique listed dependency strings, rejects unknown values and malformed inputs, and returns approval:false in every supported result. The browser uses fixed scenarios and makes no tool-specific network requests or input persistence; ordinary site analytics are separate.

Download planner.mjs, results.json and reproduce.mjs into one folder and run `node reproduce.mjs`. The fourteen results reproduce the declared mapping, not model performance. The SHA-256 in results.json identifies source bytes; it does not authenticate a publisher or prove a test date. policy.json exposes the editorial map. scenarios.json supplies all invented examples. validity-card-template.json is blank: null means no evidence supplied.

Falsification: identify a supported change that the code maps inconsistently with policy.json, or demonstrate a relevant omitted dependency. The former is an implementation error; the latter is a limitation of this illustrative policy and requires review, not an automatic accusation against a vendor. Retest scope should be designed for the actual claim and risk.

CC BY 4.0 for DN-authored files. Preserve attribution and limitations. Third-party documentation retains its terms. Corrections: https://decentralised.news/contact . Material changes require a new version.