mjolnir-qa 1.0.1 → 1.0.2

package/CHANGELOG.md

@@ -9,6 +9,12 @@ Rule behavior changes (new rules, FP-rate changes against the corpus,
 severity changes) are first-class entries here — rule IDs are immutable
 once shipped, so this file is the record of what changed between versions.
 
+## [1.0.2] — 2026-09-09
+
+### Changes since 1
+
+- P8: depth adjudication — no unexplained depth (all 89 LEXICAL rules + matrix completion + migration deferral) (#70)
+
 ## [1.0.1] — 2026-09-09
 
 ### Changes since 0

package/dist/cli.d.mts

@@ -400,6 +400,51 @@ interface Workspace {
  *   Phase 6).
  */
 type DetectionStrategy = "LEXICAL" | "AST" | "SEMANTIC" | "QA_MODEL" | "FRAMEWORK" | "RUNTIME";
+/**
+ * Closed reason-code set for `strategyJustification` (master plan P8,
+ * plan 1788853205786 — flag 6, decision 9: "no unexplained depth").
+ * A LEXICAL rule ships only when its lexical form is JUSTIFIED by one
+ * of these codes; the depth-adjudication doc renders them.
+ */
+type StrategyReasonCode =
+/** The detection target is a config/workflow surface whose statements
+ * ARE string literals (shell in YAML, key=value) — a syntax tree adds
+ * nothing the text does not already carry. */
+"shell-string-in-config" |
+/** The detection is an exact, unambiguous runner-API token (`.only`,
+ * `test.skip`, `fit`) — lexical precision equals structural precision. */
+"exact-key-match" |
+/** The defect IS the lexical artifact (commented-out test code, a
+ * recorder's default title left committed) — the text is the finding. */
+"lexical-artifact" |
+/** The semantics live in runner behavior (retries, report generation,
+ * fixture lifecycle) that no language syntax tree represents. */
+"runner-semantic" |
+/** The defect lives in string content (selector, URL) that the
+ * code-text masking deliberately excludes from structural analysis. */
+"string-content-defect" |
+/** Absence detection over a suite/directory — there is no single
+ * syntax node; the evidence is the aggregate shape of the tree. */
+"absence-aggregate" |
+/** The variant must stay in lockstep with its family's §13.2 mandatory
+ * regex fallback — the family's structural path already carries the
+ * depth, and the lexical path is the deterministic degraded mode. */
+"family-fallback-lockstep" |
+/** Migration to AST/QA_MODEL is plausible but UNPROVEN: no measurement
+ * yet demonstrates an FP reduction. Deferred to the next measurement
+ * round by owner directive (2026-09-09); the lexical detector ships
+ * with its measured tier in the meantime. */
+"migration-deferred-next-measurement";
+interface StrategyJustification {
+  /** One of the closed codes above — the machine-readable verdict. */
+  reasonCode: StrategyReasonCode;
+  /**
+   * The human-readable derivation: WHY this code applies to THIS rule,
+   * citing the detector's actual shape. Rendered by the rule docs and
+   * the depth-adjudication doc; vague boilerplate is a review rejection.
+   */
+  detail: string;
+}
 interface RuleMeta {
   /** Frozen public API — never reused (§18.4). */
   id: string;
@@ -482,6 +527,15 @@ interface RuleMeta {
    * findings that merely weaken a single test.
    */
   suiteInvalidating?: boolean;
+  /**
+   * Depth-adjudication record (master plan P8, plan 1788853205786 —
+   * flag 6, decision 9: "no unexplained depth"). REQUIRED for every
+   * rule whose `detectionStrategy` is LEXICAL (enforced by the doctor's
+   * registry-sanity check): the reason the lexical form is the honest
+   * shipped detector, with a closed reason code and a derivation that
+   * cites the detector's actual shape. Non-LEXICAL rules may omit it.
+   */
+  strategyJustification?: StrategyJustification;
 }
 interface SourceFileContext {
   /** Repo-relative path, forward slashes. */
@@ -826,7 +880,7 @@ declare const runScan: typeof runScan$1, buildUniversalRules: typeof buildUniver
  * `scripts/sync-sarif-version.cjs` on release and guarded by
  * `tests/version-consistency.spec.ts` locally.
  */
-declare const CLI_VERSION = "1.0.1";
+declare const CLI_VERSION = "1.0.2";
 /** A usage-error detail: the offending token, when one exists. */
 interface UsageErrorDetail {
   /** The unknown flag or rejected value (e.g. `--nope`, `loud`). */