feat: green-by-skip census — skip counts in the classifier, all-skipped labeling in the paid runner

bun's 'Ran N tests' line COUNTS skipped tests, so a codex/gemini shard
whose every test self-skipped (binary absent on the runner — true of every
CI runner today) exits 0, dodges the hollow-shard guard, and reads as
coverage in the weekly census. The classifier now parses bun's ' N skip' /
' N pass' recap lines; ShardOutcome carries skippedTests; formatSummary and
the fail-closed slices report label an all-skipped pass explicitly:
'all N tests SKIPPED — verified nothing'. Status stays 'passed' (external
service availability is host state, not a repo regression) but the census
can no longer mistake absence for coverage.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

test/paid-shards.test.ts

@@ -22,6 +22,7 @@ import {
   computePaidDiffSelection,
   diffSkipDecisionForFile,
   formatSummary,
+  isAllSkippedPass,
   isPaidTestFile,
   knownTestNamesInSource,
   partitionShardsByDiffSelection,
@@ -354,3 +355,39 @@ describe('parent-side diff shard skipping', () => {
     expect(summaryExitCode(withNeverStarted)).toBe(1);
   });
 });
+
+// Green-by-skip census: "Ran N tests" counts skips, so a codex/gemini file
+// whose every test self-skipped (binary absent on the runner) exits 0 and
+// used to read as coverage in the weekly report. The census label keeps the
+// pass (service availability is host state, not a repo regression) but must
+// say the shard verified nothing.
+describe('all-skipped pass census', () => {
+  const base = { shard: 1, files: ['test/codex-e2e.test.ts'], exitCode: 0, elapsedMs: 1200, groupPid: 1 };
+
+  test('isAllSkippedPass: pass with every test skipped → true', () => {
+    expect(isAllSkippedPass({ ...base, status: 'passed', executedTests: 8, skippedTests: 8 } as ShardOutcome)).toBe(true);
+  });
+
+  test('isAllSkippedPass: real work, a failure, or no data → false', () => {
+    // one test actually ran
+    expect(isAllSkippedPass({ ...base, status: 'passed', executedTests: 8, skippedTests: 7 } as ShardOutcome)).toBe(false);
+    // zero tests: that's the hollow-shard guard's territory, not this label's
+    expect(isAllSkippedPass({ ...base, status: 'passed', executedTests: 0, skippedTests: 0 } as ShardOutcome)).toBe(false);
+    // non-pass statuses never get the label
+    expect(isAllSkippedPass({ ...base, status: 'failed', executedTests: 8, skippedTests: 8 } as ShardOutcome)).toBe(false);
+    // stream gave no counts (crash/timeout) — unknown, not all-skipped
+    expect(isAllSkippedPass({ ...base, status: 'passed', executedTests: null, skippedTests: null } as ShardOutcome)).toBe(false);
+  });
+
+  test('formatSummary labels an all-skipped pass and leaves real passes alone', () => {
+    const lines = formatSummary(summarize([
+      { ...base, status: 'passed', executedTests: 8, skippedTests: 8 } as ShardOutcome,
+      { shard: 2, files: ['test/skill-e2e-review.test.ts'], status: 'passed', exitCode: 0, elapsedMs: 900, groupPid: 2, executedTests: 3, skippedTests: 0 } as ShardOutcome,
+    ]));
+    const codexLine = lines.find((l) => l.includes('codex-e2e'));
+    const reviewLine = lines.find((l) => l.includes('skill-e2e-review'));
+    expect(codexLine).toContain('all 8 tests SKIPPED');
+    expect(codexLine).toContain('verified nothing');
+    expect(reviewLine).not.toContain('SKIPPED');
+  });
+});

test/strict-output.test.ts

@@ -89,6 +89,38 @@ describe('BunTestOutputClassifier', () => {
     expect(summary.terminalFileCounts).toEqual([2]);
     expect(strictTestExitCode(0, summary, 2)).toBe(0);
   });
+
+  // "Ran N tests" COUNTS skipped tests, so the skip/pass count lines are the
+  // only stream evidence separating verified work from green-by-skip (a
+  // codex/gemini file whose every test self-skips on a binary-less runner).
+  it('parses the skip and pass count lines from bun’s recap block', () => {
+    const c = new BunTestOutputClassifier();
+    c.write(' 1 pass\n 2 skip\n 0 fail\nRan 3 tests across 1 file. [7.00ms]\n');
+    const summary = c.end();
+    expect(summary.passedTests).toBe(1);
+    expect(summary.skippedTests).toBe(2);
+    expect(summary.terminalTestCounts).toEqual([3]);
+    // all-skipped is still exit-0 at the classifier layer — the census
+    // labeling happens in the paid runner, not here
+    expect(strictTestExitCode(0, summary, 1)).toBe(0);
+  });
+
+  it('skip/pass counts survive ANSI color and chunk shears', () => {
+    const c = new BunTestOutputClassifier();
+    c.write(' 4 pa');
+    c.write('ss\n 9 skip\n');
+    const summary = c.end();
+    expect(summary.passedTests).toBe(4);
+    expect(summary.skippedTests).toBe(9);
+  });
+
+  it('prose mentioning skip counts does not pollute the tally', () => {
+    const c = new BunTestOutputClassifier();
+    c.write('console.log said: 7 skip is what we expect later\n');
+    c.write('(fail) 3 skip handling [1.00ms]\n');
+    const summary = c.end();
+    expect(summary.skippedTests).toBe(0);
+  });
 });
 
 describe('installChildSignalForwarding — cancellation terminates the RUN', () => {