test(initializer): qualify figure and reference authoring

tests/integration/README.md

@@ -29,8 +29,12 @@ The [initializer acceptance](installed/initializer-acceptance.ts) runs
 `npm create cs-object my-report` against a temporary loopback npm registry backed
 by actual archives. It installs the Python wheel into the generated project,
 starts `npm run dev`, and checks browser edits, the output-only calculation API,
-report/PDF evidence and restart. It also replaces the starter with a copy of the
-canonical two-panel calculation. Third-party dependencies still need network
+report/PDF evidence and restart. It executes the packaged composed-report/figure recipe and follows the shipped
+reference-binding recipe with independently established intermediate and output
+values. Correct, wrong and unmatched cases, formula/text changes and explicit
+rebinding are checked before browser edits, output-only curl and same-run HTML,
+PDF and evidence. Automatic results leave visual inspection and engineering
+approval separate. It also exercises a copy of the canonical two-panel calculation. Third-party dependencies still need network
 access. This tests installation without publishing a release.
 
 The canonical two-panel calculations and metadata stay in `examples/two-panel`.

tests/integration/installed/_initializer-browser.ts

@@ -19,6 +19,7 @@ export async function exerciseBrowser(options: {
   readonly outputs: Readonly<Record<string, number>>;
   readonly reportId: string;
   readonly reportTitle: string;
+  readonly referenceStatus?: 'passed' | 'not_applicable';
 }) {
   const browser = await chromium.launch({ headless: true });
   try {
@@ -115,7 +116,10 @@ export async function exerciseBrowser(options: {
     assert.equal(await page.locator('[data-slot=sheet-overlay]').count(), 0);
     await page.locator('iframe#report').waitFor({ state: 'visible' });
     assert(await page.locator('iframe#report').getAttribute('src'));
-    assert.match(await page.locator('button#pdf').innerText(), /review pending/i);
+    assert.match(
+      await page.locator('button#pdf').innerText(),
+      /review pending/i,
+    );
     assert.equal(await page.locator('#checks').isVisible(), true);
     const initialReportText = await page
       .frameLocator('iframe#report')
@@ -152,15 +156,17 @@ export async function exerciseBrowser(options: {
     assert(reportMatch?.[1]);
     const runBase = `/api/reports/${options.reportId}/runs/${reportMatch[1]}`;
     assert.equal(evidencePath, `${runBase}/evidence.json`);
-    assert.match(
-      await page.locator('#checks').innerText(),
-      /Human inspection\s+WIP/i,
-    );
-    if (options.reportId === 'rectangle-area')
+    const checksText = await page.locator('#checks').innerText();
+    assert.match(checksText, /Human inspection\s+WIP/i);
+    if (options.referenceStatus)
       assert.match(
-        await page.locator('#checks').innerText(),
-        /Independent reference\s+WIP/i,
+        checksText,
+        options.referenceStatus === 'passed'
+          ? /Independent reference\s+Passed/i
+          : /Independent reference\s+WIP/i,
       );
+    if (options.reportId === 'rectangle-area')
+      assert.match(checksText, /Independent reference\s+WIP/i);
     await page.setViewportSize({ width: 950, height: 900 });
     assert.equal(
       await page.evaluate(
@@ -230,6 +236,11 @@ export async function exerciseBrowser(options: {
       evidence.artifacts.pdf?.sha256,
       hashBytes(readFileSync(options.downloadPath)),
     );
+    if (options.referenceStatus)
+      assert.equal(
+        evidence.checks.independentReferenceAgreement?.status,
+        options.referenceStatus,
+      );
     if (options.reportId === 'rectangle-area') {
       await page.locator('[name=width]').fill('-0');
       await page.getByRole('button', { name: 'API', exact: true }).click();
@@ -270,6 +281,7 @@ export async function exerciseBrowser(options: {
         ),
       );
     }
+    return summary;
   } finally {
     await browser.close();
   }

tests/integration/installed/_initializer-evidence.ts

@@ -125,6 +125,7 @@ export async function artifactSet(options: {
     readonly unit: string;
   }[];
   readonly referencePath?: string;
+  readonly referenceStatus?: 'passed' | 'not_applicable';
   readonly name: string;
   readonly reportId: string;
 }) {
@@ -183,7 +184,8 @@ export async function artifactSet(options: {
   assert.equal(evidence.checks.sourceToDocumentConsistency?.status, 'passed');
   assert.equal(
     evidence.checks.independentReferenceAgreement?.status,
-    options.referencePath ? 'passed' : 'not_applicable',
+    options.referenceStatus ??
+      (options.referencePath ? 'passed' : 'not_applicable'),
   );
   assert.equal(evidence.checks.documentContent?.status, 'passed');
   assert.equal(evidence.checks.rendering?.status, 'passed');

tests/integration/installed/_initializer-guidance.ts

@@ -171,4 +171,5 @@ export async function exerciseGuidance(options: {
     'data:image/svg+xml',
   ])
     assert(html.includes(expected), `Missing recipe content: ${expected}`);
+  return { trial };
 }

tests/integration/installed/_initializer-processes.ts

@@ -25,6 +25,7 @@ export type CommandRunner = (options: {
   readonly args: readonly string[];
   readonly env?: NodeJS.ProcessEnv;
   readonly timeoutMs?: number;
+  readonly expectedStatus?: number;
 }) => Promise<string>;
 
 export function createCommandRunner(
@@ -97,7 +98,7 @@ export function createCommandRunner(
               `${options.command} stdout exceeded ${maximumCapturedCharacters} characters; see ${log}`,
             ),
           );
-        if (status !== 0)
+        if (status !== (options.expectedStatus ?? 0))
           return rejectRun(
             new Error(
               `${options.command} ${options.args.join(' ')} exited ${status}; see ${log}\n${stderr}${stderrTruncated ? '\n[stderr truncated in memory; full output is in the log]' : ''}`,