refactor(ci): drop the artifact-policy changes and the narration comments

The hidden-path upload fixes and their guard move to their own PR, per review:
one rule, ten uploads, artifacts with different owners. This branch keeps the
diagnose loop's own behaviour.

Comments cut to the AGENTS.md rule added in #2087. The reader's narration is
gone; what it was explaining now lives in names and types — `MeasuredVerdict`
separates xcodebuild's three words from the two the loop derives, and
`exitContradictsMeasured` names the rule that a nonzero exit outranks a green
measured test. The shape table's per-case comments are gone too; each case's
`name` already says what it is. What remains is two constants citing measured CI
numbers, which cannot be encoded.

Author: thymikee

.github/workflows/1874-diagnose.yml

@@ -131,7 +131,6 @@ jobs:
           # so every per-iteration log this loop has ever kept was silently dropped, including on
           # the run that motivated the slow-pass capture. Same input the shared diagnostics action
           # sets, for the same reason.
-          include-hidden-files: true
           path: |
             stall-summary.txt
             .tmp/stall-evidence-*.log

.github/workflows/macos.yml

@@ -122,7 +122,6 @@ jobs:
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
         with:
           name: xctest-host-results-${{ github.run_id }}-${{ github.run_attempt }}
-          include-hidden-files: true
           path: .tmp/xctest-host
           if-no-files-found: warn
 

.github/workflows/mutation-affected.yml

@@ -81,7 +81,6 @@ jobs:
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
         with:
           name: mutation-affected-select
-          include-hidden-files: true
           path: .tmp/mutation/lane-envelope.json
           if-no-files-found: warn
 
@@ -123,7 +122,6 @@ jobs:
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
         with:
           name: mutation-affected-shard-${{ matrix.name }}
-          include-hidden-files: true
           path: |
             .tmp/mutation/mutation.json
             .tmp/mutation/mutation.html
@@ -177,7 +175,6 @@ jobs:
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
         with:
           name: mutation-affected
-          include-hidden-files: true
           path: |
             .tmp/mutation/shards
             .tmp/mutation/lane-envelope.json

.github/workflows/mutation-weekly.yml

@@ -83,7 +83,6 @@ jobs:
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
         with:
           name: mutation-shard-${{ matrix.name }}
-          include-hidden-files: true
           path: |
             .tmp/mutation/mutation.json
             .tmp/mutation/lane-envelope.json
@@ -161,7 +160,6 @@ jobs:
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
         with:
           name: mutation-decision-kernels
-          include-hidden-files: true
           path: |
             .tmp/mutation/shards
             .tmp/mutation/lane-envelope.json

.github/workflows/replays-nightly.yml

@@ -92,7 +92,6 @@ jobs:
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
         with:
           name: parser-fuzz-run-${{ github.run_id }}-${{ github.run_attempt }}
-          include-hidden-files: true
           path: .tmp/fuzz
           if-no-files-found: warn
 

.github/workflows/test-app-build-cache.yml

@@ -230,7 +230,6 @@ jobs:
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
         with:
           name: ${{ matrix.artifactName }}
-          include-hidden-files: true
           path: .tmp/test-app-artifact/binary.tar.gz
           if-no-files-found: error
           compression-level: 0 # already gzipped

.github/workflows/xctest-nightly.yml

@@ -186,6 +186,5 @@ jobs:
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
         with:
           name: xctest-nightly-results-${{ github.run_id }}-${{ github.run_attempt }}
-          include-hidden-files: true
           path: .tmp/xctest-nightly
           if-no-files-found: warn

scripts/__tests__/diagnose-1874-iteration.test.ts

@@ -12,7 +12,7 @@ const script = path.join(
 );
 const NAME = 'testBareTypeUsesTappedInputWhenSoftwareKeyboardIsHidden';
 const POLL = '[DEBUG-1874] poll t=1ms observedLen=0 expectedPrefixLen=0';
-/** A real wait brackets its polls, so cadence lines and poll count are not the same number. */
+/** A wait brackets its polls, so cadence lines and poll count differ. */
 const CADENCE = [
   '[DEBUG-1874] wait start expectedLen=17',
   POLL,
@@ -21,7 +21,7 @@ const CADENCE = [
 
 const verdictLine = (test: string, word: string) =>
   `Test Case '-[X.RunnerTests ${test}]' ${word} (1.0 seconds).`;
-/** One `type` logs three phases; only `type-all` is the dispatch the threshold judges. */
+/** One `type` logs three phases; the threshold reads `type-all`. */
 const phases = (typeAllMs: string) =>
   [
     `phase=focus durationMs=644.6`,
@@ -46,7 +46,6 @@ const SHAPES: readonly (Partial<IterationReport> & { name: string; log: string[]
       lines: ['iter=1 verdict=passed rc=0 durationMs=14146 polls=1', ...CADENCE],
     },
     {
-      // Since #2035 the test skips on an ambient software keyboard, and a skip logs no phase at all.
       name: 'skipped — an environment flip, not a stall',
       log: [verdictLine(NAME, 'skipped')],
       verdict: 'skipped',
@@ -65,8 +64,6 @@ const SHAPES: readonly (Partial<IterationReport> & { name: string; log: string[]
       verdict: 'no-result',
     },
     {
-      // `pair` mode runs testBareDelayedType… alongside, and it sorts first, so its phases and
-      // verdict land in the same log ahead of the measured test's — whose 812 ms is the answer.
       name: 'pair mode — the neighbour test logs first',
       log: [
         verdictLine('testBareDelayedTypeFailsWhenTappedInputDisappearsMidCommand', 'passed'),
@@ -78,15 +75,12 @@ const SHAPES: readonly (Partial<IterationReport> & { name: string; log: string[]
       keepEvidence: false,
     },
     {
-      // xcodebuild restarts a test after an unexpected exit, so one test can report twice.
       name: 'restarted — the final word is the outcome',
       log: [verdictLine(NAME, 'failed'), ...phases('900.0'), verdictLine(NAME, 'passed')],
       verdict: 'passed',
       keepEvidence: false,
     },
     {
-      // `pair` mode: the measured test passed, but the neighbour or the runner made xcodebuild exit
-      // nonzero. Counting that as a clean pass would both miscount it and throw its log away.
       name: 'measured test passed, xcodebuild did not',
       log: [verdictLine(NAME, 'passed'), ...phases('796.1'), ...CADENCE],
       rc: 65,
@@ -110,14 +104,10 @@ test.each(SHAPES)('reads a $name iteration', (shape) => {
   const report = readIteration(shape.log.join('\n'), NAME, 1, shape.rc ?? 0);
   expect(report.verdict).toBe(shape.verdict);
   if (shape.keepEvidence !== undefined) expect(report.keepEvidence).toBe(shape.keepEvidence);
-  // Asserting the rendered lines rather than a duration field: it is what the threshold judges
-  // and what a reader of the artifact sees, so an extraction off by a digit has to show here.
   if (shape.lines) expect(report.lines).toEqual(shape.lines);
 });
 
 test('the test name is matched literally, not as a pattern', () => {
-  // Built as a regex, `testA.B` would match this line and report it passed (CodeQL: regex
-  // injection from a command-line argument).
   const log = verdictLine('testAXB', 'passed');
   expect(readIteration(log, 'testA.B', 1, 0).verdict).toBe('no-result');
   expect(readIteration(log, 'testAXB', 1, 0).verdict).toBe('passed');

scripts/diagnose-1874-iteration.ts

@@ -1,28 +1,20 @@
-// Reads one iteration of the #1874 diagnose loop (.github/workflows/1874-diagnose.yml) and writes
-// what is worth keeping. A script rather than shell in the workflow, for the reason
-// `xctest-run-summary.ts` gives: parsing inside YAML inside shell. Four defects shipped while this
-// was shell, every one of them a pipeline treating "nothing matched" as failure under `pipefail`.
+// Reads one iteration of the #1874 diagnose loop (.github/workflows/1874-diagnose.yml).
 
 import fs from 'node:fs';
 import path from 'node:path';
 import { pathToFileURL } from 'node:url';
 
-/** Healthy iterations ran 628-1131 ms (n=50); the episode on record is 2487 ms. */
+/** Measured on CI: healthy iterations 628-1131 ms (n=50), the episode on record 2487 ms. */
 const SLOW_PASS_MS = 2000;
 const CADENCE_LIMIT = 40;
 const SUMMARY = 'stall-summary.txt';
 const EVIDENCE_DIR = '.tmp';
 
+/** xcodebuild's own word for the measured test. */
+type MeasuredVerdict = 'passed' | 'skipped' | 'failed';
+
 export type IterationReport = {
-  /**
-   * xcodebuild's own word for the measured test, plus two the loop needs itself: `no-result` when
-   * the run produced no verdict — the selection matched nothing or the bundle never ran — and
-   * `run-failed` when the measured test passed but xcodebuild still exited nonzero, which in
-   * `pair` mode means the neighbour test or the runner failed. Both are ours, not the product's,
-   * and neither is evidence about the stall.
-   */
-  readonly verdict: 'passed' | 'skipped' | 'failed' | 'no-result' | 'run-failed';
-  /** Since #2035 an absorbed episode passes, so a slow pass is evidence too. */
+  readonly verdict: MeasuredVerdict | 'no-result' | 'run-failed';
   readonly keepEvidence: boolean;
   readonly lines: readonly string[];
 };
@@ -34,8 +26,6 @@ export function readIteration(
   rc: number,
 ): IterationReport {
   const lines = log.split('\n');
-  // Substring, not a pattern built from the argument: a test name is not a regex, and treating it
-  // as one lets a metacharacter match a different test.
   const marker = `${testName}]' `;
   const word = lines
     .filter((line) => line.includes(marker))
@@ -47,8 +37,6 @@ export function readIteration(
   // it as a clean pass would both miscount it and throw the log away.
   const verdict =
     rc !== 0 && measured !== 'failed' && measured !== 'no-result' ? 'run-failed' : measured;
-  // Qualified on `type-all`: an iteration also logs `phase=focus` and `phase=total`, and the
-  // threshold is calibrated on the dispatch.
   const durationMs = Number(
     lines.flatMap((line) => /phase=type-all durationMs=(\d+)/.exec(line)?.[1] ?? []).at(-1) ?? 0,
   );
@@ -66,12 +54,15 @@ export function readIteration(
       : [
           summary,
           ...cadence.slice(0, CADENCE_LIMIT),
-          // A cap that hides how much it dropped reads as "this is all there was".
           ...(dropped > 0 ? [`… ${dropped} more DEBUG-1874 lines (full log in the artifact)`] : []),
         ],
   };
 }
 
+function isMeasuredVerdict(word: string | undefined): word is MeasuredVerdict {
+  return word === 'passed' || word === 'skipped' || word === 'failed';
+}
+
 function main(): number {
   const [logPath, testName, iteration, rc] = process.argv.slice(2);
   if (!logPath || !testName || !iteration || !rc) {