devswha · Sep 26, 2026 · Sep 26, 2026
diff --git a/artifacts/rebaseline-2025/README.md b/artifacts/rebaseline-2025/README.md
@@ -84,15 +84,6 @@ uses MIT for its own code and documentation, but that license does not relicense
 third-party web pages. Keep web extracts hash-only unless the exact row has a
 redistribution review that permits publishing the text.
 
-To run the private KatFish KO calibration after downloading `essay.jsonl`, `abstract.jsonl`, and `poetry.jsonl` into `artifacts/rebaseline-2025/private/katfish/`:
-
-```bash
-npm run benchmark:katfish-ko -- --write --basename katfish-ko-latest
-```
-
-The command writes only aggregate metrics to `docs/benchmarks/`; it must not
-commit KatFish raw rows unless a license review explicitly allows it.
-
 To refresh their deterministic score/outcome fields from a private raw-text
 intake file:
 

diff --git a/docs/ARCHITECTURE.md b/docs/ARCHITECTURE.md
@@ -256,8 +256,7 @@ The checker enforces only three reachability rules:
 2. `playground/**` cannot reach server-secret modules, API handlers, or Node
    built-ins.
 3. `src/**`, `api/**`, `bin/**`, and every declared published package bin
-   cannot reach research modules (`scripts/research/`, `tests/quality/`, and
-   `scripts/iterative-rewrite-baseline.mjs`).
+   cannot reach research modules (`scripts/research/` and `tests/quality/`).
 
 Deterministic shared modules are classified rather than blanket-banned:
 `src/edit-controls.js`, `src/errors.js`, `src/logger.js`,

diff --git a/docs/COOKBOOK.md b/docs/COOKBOOK.md
@@ -75,11 +75,7 @@ The Action finds changed Markdown files with `dorny/paths-filter@v4`, runs patin
 
 ### README score badge
 
-`npm run badge` writes a [Shields.io endpoint](https://shields.io/endpoint) JSON file from the same deterministic prose score. It reports the highest scored file (`maxScore`) as an editing-hotspot percentage, not an authorship verdict.
-
-```bash
-npm run badge -- README.md docs/FAQ.md > patina-badge.json
-```
+The Action's `patina-badge.json` is a [Shields.io endpoint](https://shields.io/endpoint) file built from the same deterministic prose score. It reports the highest scored file (`maxScore`) as an editing-hotspot percentage, not an authorship verdict.
 
 ```json
 { "schemaVersion": 1, "label": "patina", "message": "25% · human-ish", "color": "brightgreen" }

diff --git a/docs/HARNESS.md b/docs/HARNESS.md
@@ -36,9 +36,9 @@ INPUT          ENGINE (deterministic, LLM-free)                 SURFACES
   │ HARNESS                                                                    │
   │  (1) regression (det.)   benchmark · benchmark:ranges · benchmark:report   │
   │  (2) calibration (det.)  signal-impact · rebaseline:score/report ·         │
-  │                          low-fpr · katfish-ko · lexicon:freshness          │
+  │                          low-fpr · lexicon:freshness                       │
   │  (3) robustness/perf     robustness · perf                                 │
-  │  (4) LLM quality (opt-in) quality:live · quality:rewrite-ab · adversarial-mps │
+  │  (4) LLM quality (opt-in) quality:live · adversarial-mps                   │
   │  (5) comparison (det.)   detector-comparison                               │
   └──────────────────────────────────────────────────────────────────────────┘
                        │  enforced by
@@ -72,7 +72,6 @@ analysis below ablates each one to report its marginal contribution.
 | Rebaseline summary | `npm run benchmark:rebaseline` / `:report` | Validates the public manifest + computes the headline catch/FP claim | [tests/quality/README.md](../tests/quality/README.md) |
 | Rebaseline score | `npm run benchmark:rebaseline:score` | Re-runs the analyzer over private rows → public scored manifest (predicted_hot + trigger_counts) | [tests/quality/README.md](../tests/quality/README.md) |
 | Low-FPR metrics | `npm run benchmark:rebaseline:low-fpr` | TPR@1%/5%FPR operating points | `scripts/rebaseline-low-fpr-report.mjs` |
-| KatFish calibration | `npm run benchmark:katfish-ko` | KO diagnostic catch/FP deltas (aggregate-only) | [tests/quality/README.md](../tests/quality/README.md) |
 | Lexicon freshness | `npm run lexicon:freshness` | AI-lexicon provenance sidecar check | `scripts/lexicon-freshness.mjs` |
 | FP fixture export | `node scripts/fp-fixture-export.mjs` | Turns FP reports into suspect-zone fixtures | `scripts/fp-fixture-export.mjs` |
 
@@ -82,7 +81,6 @@ analysis below ablates each one to report its marginal contribution.
 |---|---|---|---|
 | Live rewrite quality | `npm run quality:live` (`PATINA_LIVE=1` to call a model) | before/after AI score, MPS, fidelity on rewrites | [tests/quality/README.md](../tests/quality/README.md) |
 | Adversarial MPS | `npm run quality:adversarial-mps` | Guards against MPS hiding unchanged AI style | [tests/quality/README.md](../tests/quality/README.md) |
-| Rewrite A/B | `npm run quality:rewrite-ab` (`--live`) | Compares two rewrite configs (default `single` vs `iterative-baseline`) on the same fixtures: after-AI/MPS/fidelity/edit-churn + per-fixture winner. Answers whether the baseline comparison flow produces a better rewrite. | [tests/quality/README.md](../tests/quality/README.md) |
 
 ## Gates (deterministic, run in CI / pre-publish)
 
@@ -92,7 +90,7 @@ analysis below ablates each one to report its marginal contribution.
 | Tests | `npm test` | unit + e2e (`node --test`) |
 | Release metadata | `npm run release:check` | version sync across all version-bearing surfaces |
 | Private-asset leak | `npm run check:no-private-assets` | no private/vendor text in the npm tarball or tree |
-| Prose score gate | `patina-score` / `npm run badge` | hot-paragraph ratio CI gate (default 30) |
+| Prose score gate | `patina-score` | hot-paragraph ratio CI gate (default 30) |
 
 ## Signal impact harness (`scripts/signal-impact.mjs`)
 

diff --git a/docs/QA.md b/docs/QA.md
@@ -37,11 +37,10 @@ result `inconclusive` until an approved command exists.
 | `performance-package-size` | `npm run benchmark:perf -- --costmetrics` | Explicit report-only tarball size, cold-process CLI timing, and warm analyzer measurements; no rebaseline or latency gate |
 | `dogfood` | `npm run dogfood` | Configured public-document checks |
 | `release-safety` | `npm run release:check` and `npm run check:no-private-assets` | Release metadata and private-asset boundary |
-| `mock-quality` | `npm run qa:mdx` | Existing MDX QA script, when its scope applies |
 | `live-model` | `npm run quality:live` | Opt-in model-backed quality; approval and budget required |
 
 Other package scripts may fit the changed boundary (for example,
-`benchmark:robustness`, `quality:adversarial-mps`, or `quality:rewrite-ab`).
+`benchmark:robustness` or `quality:adversarial-mps`).
 Record the exact command and arguments rather than assuming a similarly named
 script exists. `benchmark:rebaseline*` and other research commands are not
 routine acceptance profiles.

diff --git a/package.json b/package.json
@@ -35,7 +35,6 @@
     "quality:adversarial-mps": "node scripts/adversarial-mps-report.mjs --check",
     "quality:persona-ablation": "node scripts/persona-ablation.mjs",
     "dogfood": "node tests/quality/dogfood.mjs",
-    "badge": "node scripts/badge-json.mjs",
     "card": "node scripts/share-card.mjs",
     "lint": "npm run lint:syntax && npm run lint:eslint && npm run typecheck && npm run spellcheck && npm run check:architecture",
     "release:check": "node scripts/check-release-metadata.mjs && node scripts/check-retired-concepts.mjs",
@@ -49,16 +48,13 @@
     "lint:eslint": "eslint .",
     "typecheck": "tsc -p .",
     "spellcheck": "cspell \"patterns/**/*.md\" \"lexicon/**/*.md\" --gitignore-root .",
-    "qa:mdx": "node scripts/qa/mdx-score.mjs",
     "quality:live": "node tests/quality/live-quality.mjs",
     "benchmark:scorer:live": "node tests/quality/live-scorer-benchmark.mjs",
     "quality:model-rewrite": "node scripts/research/model-rewrite-benchmark.mjs",
-    "quality:rewrite-ab": "node scripts/rewrite-ab.mjs",
     "benchmark:rebaseline:intake": "node scripts/rebaseline-intake.mjs",
     "benchmark:rebaseline:score": "node scripts/rebaseline-score.mjs",
     "benchmark:signal-impact": "node scripts/signal-impact.mjs",
     "benchmark:rebaseline:web": "node scripts/rebaseline-web-collect.mjs",
-    "benchmark:katfish-ko": "node scripts/katfish-calibration.mjs",
     "lexicon:freshness": "node scripts/lexicon-freshness.mjs --check",
     "benchmark:rebaseline:generate-modern": "node scripts/rebaseline-generate-modern.mjs",
     "benchmark:rebaseline:claim-manifest": "node scripts/rebaseline-build-claim-manifest.mjs",

diff --git a/scripts/badge-json.mjs b/scripts/badge-json.mjs
diff --git a/scripts/check-architecture.mjs b/scripts/check-architecture.mjs
@@ -157,7 +157,6 @@ const SERVER_SECRET_MODULES = Object.freeze([
 ]);
 
 const RESEARCH_MODULES = Object.freeze([
-  'scripts/iterative-rewrite-baseline.mjs',
   'scripts/research/',
   'tests/quality/',
 ]);