Offline: which deployment an advisory may affect, what to retest after a change, whether a fix was verified.
keryn revalidate plan <input.json> [--out <file>]Options
Section titled “Options”Usage: keryn revalidate inputs <result.json> <binding.json> [--out <file>] keryn revalidate affectedness <inventory.json> <affected.json> [<observation.json>...] --now <iso> [--vex <file>]... [--out <file>] keryn revalidate plan <input.json> [--out <file>] keryn revalidate apply <plan.json> <results.json> [--out <file>] keryn revalidate intervention <input.json> --now <iso> [--out <file>] Offline and read-only: every subcommand reads JSON files and writes JSON. Nothing is scanned, sent, executed or approved, and no release decision changes. inputs the inventory and the affected packages of a Keryn result JSON (its locked packages and dependency findings), bound to a build by <binding.json>: {tenant, product, build: {sha256, revision}, origin}. A package with no exact identity is listed as unresolved and left out. affectedness accepts this file in place of the two. affectedness advisory -> component -> build -> observed deployment (keryn.dev/affectedness/v1). The inventory is keryn.dev/build-inventory/v1 (tenant, product, BOM and build digests, components by purl); each observation holds a keryn.dev/deployed-assets/v1 receipt. Every edge is exact, inferred, ambiguous, contradiction or unknown, with the reason; a near identity is listed and never matched. An observation older than 72 hours is not used. --vex adds supplier statements as attributed claims; they change no edge. plan {tenant, events, subjects, providers} -> keryn.dev/retest-plan/v1: the checks to run again after a changed advisory, build or deployment binding, or authority. It authorizes nothing; an affected set that cannot be bounded gives explicit full rechecks. apply which planned subjects were really retested. A carried result, a result from before the plan or another check version does not count. Results: {plannedAt, results}. intervention {record, baseline, after, rollback, retest, alternativePaths} -> keryn.dev/intervention-verdict/v1: still observed, verified or mitigated within the tested scope, not comparable, inconclusive, or reappeared. In place of baseline and after, ai: {baselineSuite, baselineReceipt, currentSuite, currentReceipt, context} has the two runs rebuilt from AI test suites and receipts by the existing fix verification. Exit 0: computed and complete. 3: the result says it is incomplete (a plan that could not bound an affected set, a subject still stale, a verdict that is inconclusive or not comparable). 2: unreadable or malformed input; 4 internal.Examples
Section titled “Examples”# Plan only; running the checks is a separate, authorized step.keryn revalidate plan changes.json --out retest-plan.jsonExit codes
Section titled “Exit codes”| Code | Meaning |
|---|---|
| 0 | Computed, and the result is complete. |
| 2 | Unreadable or malformed input. |
| 3 | Computed, and the result says it is incomplete: a plan that could not bound an affected set, a subject still stale, a verdict that is inconclusive or not comparable. |
| 4 | Internal error. |