# MOCK-ENGAGEMENT-1 — Engagement Package for Review (generated; NOT published)

**COVER-GEN-1:** this cover is GENERATED from the package artifacts by
`../engagement-corpora/engagement-package-check.mjs` — the SHIPPED copy of
`scripts/engagement-package-check.mjs` (byte-identical; REPRO-PATH-1, R6-8:
the generator is listed in the package and named package-relative, never
repo-relative; Tani 2026-09-14T06:44Z ruling item
10) — never hand-typed. Every number below is a read-back from
`findings/findings.json`, `findings/closure-receipts.json`, and
`findings/engagement-labor-ledger.json` at generation time. DETERMINISTIC:
no wall-clock input — the meaningful dates below (eval_at, the ran window)
are read back from findings.json; identical artifacts regenerate identical
cover bytes (a hand edit cannot survive, and a check run cannot dirty the
tree).

SYNTHETIC — MOCK ENGAGEMENT. The buyer is invented; every artifact of this engagement carries that label. Real provider rows carry ONLY real board/NPPES facts; the wrong values live ONLY on synthetic rows (invented persons).

Buyer: "Example Health Plan (SYNTHETIC)" · engagement:
"MOCK-ENGAGEMENT-1" · eval_at
2026-09-14T20:51:58.210Z · ran started 2026-09-16T07:35:48.798Z / finished 2026-09-16T07:35:48.824Z.

## Reading order — as the buyer's reviewer would read it

1. `../engagement-scope.md` — the commissioned question, field scope, exclusions.
2. `../roster.json` + `../sample-freeze.json` — the frozen inputs (25 frozen row selection(s) over a 60-row roster). SAMPLE-COMMIT-1: the pre-run commitment is git commit a2031937 (2026-09-13T14:38:33Z; tree at that commit = scope + roster + sample-freeze only, no findings machinery — verified read-back); roster sha256 `91e96b6a59b7bea5…` + ordered sampling-frame sha256 `5d0fb34f8078c66c…` are recorded in sample-freeze.json and ENFORCED by leg [12] — the claim is bounded to what that git commitment and these enforced digests establish.
3. `findings/ADJUDICATION-LAW.md` — the published comparison laws, all
   versions (a reviewer may dispute any rule). Current law: ADJUDICATION-LAW-v8 (STATE-ALIAS-1 + CROSS-STATE-1 + LAW-IN-THE-BRANCH-1, Tani 2026-09-15T04:30Z — REPRO-DRIFT-1 gate + R7-8 rerun + forbidden assertions enforced in the emitting branch).
4. `findings/findings.json` — the signed bundle (current law; carries the
   full v1→v2→v3→v4 computed delta).
5. `findings/closure-receipts.json` — CLOSURE-RECEIPT-v1 §4-shaped; routes
   assigned, never closed (41/41 carry closure_route_state ASSIGNED — validated before this cover asserts it); null authorization WITH REASON (synthetic buyer).
6. `findings/engagement-labor-ledger.json` — per-phase effort, basis line
   per number.
7. `findings/SHA256SUMS` — and the engagement public key ships ONLY inside
   findings.json (trust.public_key; KEY-IN-TREE-1: key material lives outside
   the tree — no .pem under public/).
8. `../fixture-authorities.json` — the SYNTHETIC FIXTURE AUTHORITY corpus
   for the five invented providers (MOCK-2) — never the pinned real corpora.

## The honest counts — read back from findings.json

Current state (100 propositions over 25 rows):

| state | rows-propositions |
|---|---|
| VERIFIED | 59 |
| CONTESTED | 23 |
| UNVERIFIABLE | 17 |
| STALE | 1 |

Across law versions (v1 → v2 NORMALIZATION-1 → v3 PARSE-ADDRESS-1 → v4 ADJUDICATION-2 → v5 NPI-ROLE-1 → v6 DETERMINACY-1 → v7 STATE-ALIAS-1/CROSS-STATE-1 → v8 LAW-IN-THE-BRANCH-1):

| state | v1 | v2 | v3 | v4 | v5 | v6 | v7 | v8 (current) |
|---|---|---|---|---|---|---|---|---|
| VERIFIED | 48 | 48 | 48 | 59 | 59 | 55 | 59 | 59 |
| CONTESTED | 32 | 32 | 32 | 22 | 22 | 26 | 24 | 23 |
| UNVERIFIABLE | 20 | 20 | 20 | 18 | 18 | 18 | 16 | 17 |
| STALE | 0 | 0 | 0 | 1 | 1 | 1 | 1 | 1 |

Computed flips: v1→v2 0 · v2→v3 0 · v3→v4 34 · v4→v5 0 · v5→v6 4 · v6→v7 6 · v7→v8 1
(every flip computed by the run, never asserted — and every delta RECONSTRUCTS its own destination: applying the published flips to the published from-counts lands on the published to-counts, verified newest-first by package-check leg [17], DELTA-ARITHMETIC-1).

**Our own finding during R8 (not the reviewer's catch):** the v7-signed bundle asserted the v6-forbidden "two different published persons" phrase on FIVE rows (ROW-020, ROW-023, ROW-058 and two more) — found by our own gate leg [16] applied to our own signed bundle, the first application of CLASS-NOT-VECTOR-1 (standing law, Tani 2026-09-15T04:30Z): the reported vector was one row; the class was five.

## Synthetic demonstration (MOCK-2) — property stated exactly

"none of the labeled bad fields was VERIFIED"

"MOCK-2 (Tani 2026-09-14T06:44Z ruling item 8): each synthetic buyer row carries exactly ONE labeled defect on an otherwise valid fixture baseline, adjudicated against separately labeled SYNTHETIC FIXTURE authorities (fixture-authorities.json — never the pinned real corpora). Demonstrated property, stated exactly (R6-3 COVER-ASSERT-1, Tani 2026-09-14T18:08Z: the mechanism-detection claim is struck back to the demonstrated property): none of the labeled bad fields was VERIFIED; unaffected fields retained their supported outcomes; dependent fields blocked by a labeled defect are disclosed per row and never counted as detected defects. Never a mechanisms count."

Labeled synthetic rows in the demonstration: ROW-056, ROW-057, ROW-058, ROW-059, ROW-060.

## Closure receipts — ownership tallies, read back

41 receipts; 41/41 carry closure_route_state ASSIGNED — routes are assigned, never closed (closure is not expressible on a receipt; FOLLOW-UP-DEF-1 term 1; validated before assertion, COVER-ASSERT-1).

| gap_owner / closure_route | count |
|---|---|
| `PUBLISHER/REMEDIATED_PUBLICATION` | 4 |
| `REGISTRY/ACCEPTED_LIMITATION` | 1 |
| `REGISTRY/REGISTRY_INVESTIGATION` | 3 |
| `UNRESOLVED-BINDING/BINDING_RESOLUTION` | 19 |
| `WOC-METHOD/FIXTURE-BY-DESIGN` | 9 |
| `WOC-METHOD/REMEDIATE_METHOD` | 5 |

## The labor ledger — real minutes, read back from the file

**Totals, read back after the write: 246 agent-minutes /
31 human-minutes** across 10 phases:

  - preparation: 24 agent / 0 human
  - acquisition: 18 agent / 5 human
  - sampling-freeze: 9 agent / 0 human
  - adjudication: 38 agent / 0 human
  - delivery: 31 agent / 0 human
  - NORMALIZATION-1 rework: 49 agent / 11 human
  - PARSE-ADDRESS-1 rework: 35 agent / 8 human
  - INSTRUMENT-1 rework: 16 agent / 3 human
  - ADJUDICATION-2 rework (R5-1): 26 agent / 4 human
  - follow-up: 0 agent / 0 human

**REPORT-FILE-DIVERGENCE-1 (standing, ruled by Tani 2026-09-13T19:07Z):** no
number reaches Tani unless it was read back from the artifact after the
write. This cover is that law applied to itself: the check that generates it
is the same check that fails the package on a diverging artifact.

## Reproduce the engagement package (REPRO-PATH-1, R6-8 — one package-relative command, fixed inputs, output OUTSIDE the frozen delivery)

```
# a clean offline copy of this package — the copy is not the frozen delivery:
cp -r . /tmp/r6-repro && cd /tmp/r6-repro
node engagement-corpora/engagement-findings.mjs --out engagement-mock/findings
```

The shipped machinery resolves the corpora and the frozen inputs beside
itself; the clean copy alone — manifest-listed inputs only — regenerates the
package deterministically (except eval_at/timestamps), runs the shipped
package check, and re-generates this cover from the artifacts. No private
files, no repo, no network. The shipped machinery
(engagement-corpora/engagement-findings.mjs + engagement-corpora/engagement-package-check.mjs)
is byte-identical to scripts/engagement-findings.mjs / scripts/engagement-package-check.mjs.