paperlint 3.2.0 → 4.1.0

package/skills/paper-pipeline/scripts/pipeline-check.harness.mjs

@@ -1,328 +0,0 @@
-/**
- * pipeline-check.harness.mjs — one planted defect at a time, on an otherwise-clean scorecard.
- *
- * 🔴 Since 2026-08-26 HERE ARE SIX CHECKS, NOT TWENTY-TWO. Sixteen, whose input is the document itself,
- * moved to `eslint-rules/pipeline-status.mjs`, and their cases — to `eslint-rules/pipeline-status.harness.mjs`.
- * `npx vigiles test .claude/skills/paper-pipeline/scripts/pipeline-check.harness.mjs`.
- *
- * WHY THIS EXISTS ALONGSIDE THE FIXTURE TEST IN hooks.harness.ts. That one runs the checker over
- * `scripts/fixtures/dirty/`, a scorecard carrying every defect at once, and asserts each kind
- * appears somewhere in the output. It cannot answer the question that matters when a check breaks:
- * WHICH defect produced WHICH finding. A checker that reports `stale-continuous` for the wrong
- * reason — say, because the date parser returns null and every row looks undated — passes that test
- * forever. So each case below plants exactly ONE defect on a scorecard that is otherwise silent, and
- * asserts the FULL SET of kinds, not merely that the expected one is present. Collateral findings
- * are written into the expectation on purpose: they are part of the contract, and an unlisted one is
- * a change in behaviour nobody asked for.
- *
- * 🔴 Assertions run at MODULE TOP LEVEL. `vigiles test` imports the file and treats "did not throw"
- * as a pass — an earlier harness here exported a `tests` object, nothing ran, and the runner printed
- * ✓ on a file whose only assertion was `assert.equal(1, 2)`.
- *
- * 🔴 The subprocess environment is SCRUBBED of OSF_TOKEN and GITHUB_TOKEN. The `credential-available`
- * check fires when the session holds a credential that could clear a parked blocker, so with either
- * token exported this suite would pass on a laptop and fail in CI, or the reverse. A test whose
- * verdict depends on who is running it is not a test.
- *
- * Fixtures are throwaways in a temp dir; no real scorecard is read or written.
- */
-import assert from "node:assert/strict";
-import { parseStatus } from "./pipeline-check.mjs";
-import { execFileSync } from "node:child_process";
-import {
-  mkdtempSync,
-  mkdirSync,
-  writeFileSync,
-  rmSync,
-  realpathSync,
-} from "node:fs";
-import { tmpdir } from "node:os";
-import { join, dirname } from "node:path";
-import { fileURLToPath } from "node:url";
-import { consumerRoot } from "./consumer.mjs";
-import { scorecard as makeScorecard } from "./fixtures/scorecard.ts";
-const HERE = dirname(fileURLToPath(import.meta.url));
-
-const ROOT = consumerRoot();
-const SCRIPT = join(HERE, "pipeline-check.mjs");
-const tmp = realpathSync(mkdtempSync(join(tmpdir(), "pipecheck-harness-")));
-
-// The paper source is written NOW, so `newestSourceDate` returns today and the staleness comparison
-// has something real to compare against. Every clean row therefore carries today's date: hard-coding
-// a date would make this suite start failing on its own the following morning.
-const TODAY = new Date().toISOString().slice(0, 10);
-
-const ENV = { ...process.env };
-delete ENV.OSF_TOKEN;
-delete ENV.GITHUB_TOKEN;
-
-const scorecard = (over = {}) => makeScorecard(over, TODAY);
-
-/** Write a fixture and return the findings pipeline-check reports, verbatim. */
-function findings(name, over = {}, today = TODAY) {
-  const d = join(tmp, name);
-  mkdirSync(d, { recursive: true });
-  writeFileSync(join(d, "PIPELINE-STATUS.md"), scorecard(over));
-  writeFileSync(join(d, "paper.md"), "## 1. Heading\n\nProse.\n");
-  const out = execFileSync("node", [SCRIPT, d, "--json", `--today=${today}`], {
-    cwd: ROOT,
-    encoding: "utf8",
-    env: ENV,
-    stdio: ["ignore", "pipe", "pipe"],
-  });
-  return JSON.parse(out);
-}
-
-/** …and just the kinds, sorted. */
-const kinds = (name, over = {}, today = TODAY) =>
-  findings(name, over, today)
-    .map((f) => f.kind)
-    .sort();
-
-/** A SETUP-shaped (7-column) table head. */
-const PLAIN_HEAD = [
-  "| id | Work | Skill | Status | Date | Result | Open |",
-  "|----|------|-------|--------|------|--------|------|",
-];
-
-/** Assert the EXACT set of kinds, so a check firing for the wrong reason cannot hide. */
-const expect = (name, over, want, today = TODAY) =>
-  assert.deepEqual(
-    kinds(name, over, today),
-    [...want].sort(),
-    `${name}: wrong finding set`,
-  );
-
-// ═══════════════════════════════════════════════════════════════════════════════════════
-// 🔴 2026-08-26 — SIXTEEN CASES MOVED OUT ALONG WITH THEIR CHECKS.
-//
-// Checks whose input is ONE markdown document moved to `eslint-rules/pipeline-status.mjs`,
-// and their cases — to the colocated `eslint-rules/pipeline-status.harness.mjs`. They were NOT
-// left behind even "just in case": a test running a check that is not in the file is always green
-// and therefore lies worse than its absence.
-//
-//   undated-continuous · gate-missing-input · gate-stale-input · unknown-input ·
-//   unattributed-verdict · single-family-jury (+ mixed-family / non-judging / widened) ·
-//   scalar-verdict (+ owned) · study-before-frame · undeclared-input (+ superset) ·
-//   unrun-gate · no-verdict · weak-accept-unowned · ceiling-unplanned ·
-//   unattributed-row · unreadable-row · duplicate-id
-//
-// Six checks remain, and all six read input FROM OUTSIDE the file: git, clock, `reviews/`,
-// `process.env`. Exactly those are checked below.
-//
-// ⚠️ TWO CASES ABOUT PARSING WERE REWRITTEN, NOT DELETED, and this is more important than keeping them.
-// "Heading with trailing text" and "line `| Id |`" — parser regressions, found live
-// on 2026-08-09, and before they were observed through `unrun-gate`, which no longer exists here. If they
-// had simply been left expecting `[]`, they would have turned green BY CONSTRUCTION: empty set
-// of findings appears both when the parser is right and when it silently ate the table. So both
-// were rewritten to be observed through `stale-continuous` — the check that stayed here:
-// parser did not see section ⇒ no line ⇒ no finding ⇒ assertion is red.
-// ═══════════════════════════════════════════════════════════════════════════════════════
-
-// ── 0. the baseline says nothing ───────────────────────────────────────────────────────
-// Asserted first and asserted hard. Every case below is "the baseline plus one thing", so a noisy
-// baseline would make all of them meaningless — and a checker that fires on a clean scorecard is
-// muted within a day, which is worse than one that misses.
-expect("clean", {}, []);
-
-// ── 1. stale-continuous — a ☑ about text that has since been rewritten ─────────────────
-// verify-citations was numbered like a one-shot step, so a citation added after the last run counted
-// as verified. The project rule "we checked last cycle does not count" was already written down.
-// INPUT FROM OUTSIDE THE FILE: the date of the last commit that touched the paper's text — that is why the check stayed.
-expect("stale-continuous", { cites: { date: "2020-01-01" } }, [
-  "stale-continuous",
-]);
-
-// ── 2. submit-access — no portal account inside somebody else's moderation window ──────
-// Found at T−4 days on a finished paper, as the sole item on the critical path: OpenReview profiles
-// are moderated for up to two weeks and there is no expedite route without an institutional email.
-// INPUT FROM OUTSIDE THE FILE: the clock.
-expect(
-  "submit-access",
-  { access: { status: "☐" }, deadline: "2026-09-01" },
-  ["submit-access"],
-  "2026-08-20",
-);
-
-// ── 3. …and a far deadline with the same blank row is NOT urgent ───────────────────────
-expect("submit-access-not-yet", { access: { status: "☐" } }, []);
-
-// ── 4. a history section is evidence, not status ──────────────────────────────────────
-// A superseded 2026-08-02 row saying the artifact was unhosted re-triggered the credential check on
-// 2026-08-05, hours after the artifact went up. This project appends rather than rewrites, so every
-// scorecard accumulates old rows, and parsing them makes every past state live again.
-// Observed through `stale-continuous`: a row in history is dated 2020, and if the history
-// slice did not work, it would have given it.
-expect(
-  "history-is-not-status",
-  {
-    tail: [
-      "## History",
-      "",
-      "### CONTINUOUS",
-      ...PLAIN_HEAD,
-      "| cites | any cite | verify-citations | ☑ | 2020-01-01 | old | — |",
-    ].join("\n"),
-  },
-  [],
-);
-
-// ── 5. a heading may carry trailing text — the RELAXATION, still observable ────────────
-// The old pattern was `/^###\s+([A-Z]+)\s*$/`: an all-caps Latin word and NOTHING after it. Papers
-// annotate their headings (`### GATES (re-run 06.08)`), so the pattern failed on the real document
-// and the section simply stayed unset — SILENTLY. Planted here as a CONTINUOUS row with an old date:
-// the finding exists only if the annotated heading was recognised as a section.
-expect(
-  "heading-with-trailing-text",
-  {
-    tail: [
-      "### CONTINUOUS (re-run 2026-08-09, after the cut)",
-      ...PLAIN_HEAD,
-      "| refs | any ref moved | verify-refs | ☑ | 2020-01-01 | ok | — |",
-    ].join("\n"),
-  },
-  ["stale-continuous"],
-);
-
-// ── 6. the row `| Id |` is a ROW, and the header is whatever `thead` says it is ────────
-// The old parser guessed the header with `/^\|\s*id\s*\|/i`, so the DATA row `| Id | …` — the
-// research-ideate row, before the words-not-codes rename — matched the guess and was eaten on every
-// scorecard in the repo, for as long as the template had existed. A row that parses as nothing, on
-// every paper, forever. Planted as a stale CONTINUOUS row so its absence is visible: if the row were
-// still being swallowed, the finding set would be EMPTY.
-{
-  const f = findings("header-lookalike-row", {
-    tail: [
-      "### CONTINUOUS",
-      ...PLAIN_HEAD,
-      "| Id | Header-lookalike | research-ideate | ☑ | 2020-01-01 | — | — |",
-    ].join("\n"),
-  });
-  assert.deepEqual(
-    f.map((x) => x.kind),
-    ["stale-continuous"],
-    "header-lookalike-row: wrong finding set",
-  );
-  assert.match(
-    f[0].msg,
-    /\bId\b/,
-    "header-lookalike-row: the swallowed row is not named in the finding",
-  );
-}
-
-// ── 17. the cold-read report's filename: BOTH live spellings must count ───────────────────────────
-// 2026-08-24: this gate announced "nobody without context has read this text" about a paper that had
-// been cold-read TWICE that day. The report was filed as `…-coldread-worksheet.md`; the pattern knew
-// only `cold-read`. Neither spelling is a typo — the skill is `cold-read-diff` while the scorecard
-// row id, in the very file this checker parses, is `coldread` — so a gate keyed to one of two live
-// spellings is guaranteed to miss. Its failure mode is the expensive one: a hole announced where
-// none exists is how a gate stops being read, and then a real hole walks past it.
-for (const fname of [
-  "2026-08-24-cold-read-report.md",
-  "2026-08-24-coldread-worksheet.md",
-]) {
-  const name = "coldread-" + fname.replace(/[^a-z]/gi, "");
-  const d = join(tmp, name);
-  mkdirSync(join(d, "reviews"), { recursive: true });
-  writeFileSync(join(d, "PIPELINE-STATUS.md"), scorecard());
-  writeFileSync(join(d, "paper.md"), "## 1. Heading\n\nProse.\n");
-  writeFileSync(join(d, "reviews", fname), "# a reader with no context\n");
-  const out = execFileSync("node", [SCRIPT, d, "--json", `--today=${TODAY}`], {
-    cwd: ROOT,
-    encoding: "utf8",
-    env: ENV,
-    stdio: ["ignore", "pipe", "pipe"],
-  });
-  const ks = JSON.parse(out).map((f) => f.kind);
-  assert.ok(
-    !ks.includes("no-cold-read"),
-    `a real cold-read report named ${fname} was reported as missing — the pattern does not know this spelling`,
-  );
-}
-// …and the half that proves the check still works: reviews/ exists and holds nothing cold-read.
-{
-  const d = join(tmp, "coldread-absent");
-  mkdirSync(join(d, "reviews"), { recursive: true });
-  writeFileSync(join(d, "PIPELINE-STATUS.md"), scorecard());
-  writeFileSync(join(d, "paper.md"), "## 1. Heading\n\nProse.\n");
-  writeFileSync(
-    join(d, "reviews", "2026-08-24-hotcrp-reviews.md"),
-    "# not a cold read\n",
-  );
-  const out = execFileSync("node", [SCRIPT, d, "--json", `--today=${TODAY}`], {
-    cwd: ROOT,
-    encoding: "utf8",
-    env: ENV,
-    stdio: ["ignore", "pipe", "pipe"],
-  });
-  assert.ok(
-    JSON.parse(out)
-      .map((f) => f.kind)
-      .includes("no-cold-read"),
-    "reviews/ with no cold-read report was NOT reported — widening the pattern silenced the gate",
-  );
-}
-
-rmSync(tmp, { recursive: true, force: true });
-console.log(
-  "✓ pipeline-check: 6 cases + two spellings of the cold-read report — one planted defect at a time, on six checks whose input is OUTSIDE the file (git · clock · reviews/ · env). The other sixteen moved to eslint-rules/pipeline-status.harness.mjs",
-);
-
-// ── THE HISTORY CUT ──────────────────────────────────────────────────────────
-// Added 2026-09-19, and the reason it was added is the finding: `currentPart` had NO test at
-// all. The word "History" appeared nowhere in this file, so the function that decides which
-// half of a scorecard counts as current was running unwatched.
-//
-// 🔴 That is also why the cleanup it guards was unsafe until now. `HISTORY_MARK_RE` — a regex
-// over the raw sentence «Всё, что ниже, — история» — was deleted in favour of the heading path
-// alone, and neither the old branch nor the new one could be shown to work. Two attempts at a
-// discriminator through the CLI returned the same answer for both inputs, which says nothing
-// about the subject: same answer for different inputs means the probe is measuring itself.
-// Going through the exported parser instead separates them in one call.
-//
-// Both halves: the buried row is gone under a marker heading, and PRESENT under one that is not
-// a marker. Without the second, a parser that dropped every second section would pass.
-{
-  const body = (heading) => `# Scorecard
-
-### CONTINUOUS
-| id | Trigger | Skill | Status | Date | Result | Open |
-|----|---------|-------|--------|------|--------|------|
-| live | x | s | ☑ | 2026-09-01 | current | — |
-
-## ${heading}
-
-### CONTINUOUS
-| id | Trigger | Skill | Status | Date | Result | Open |
-|----|---------|-------|--------|------|--------|------|
-| buried | x | s | ☑ | 2001-01-01 | superseded snapshot | — |
-`;
-  const ids = (t) =>
-    Object.values(parseStatus(t).sections)
-      .flat()
-      .map((r) => r.id);
-
-  for (const marker of [
-    "History",
-    "Archive",
-    "history",
-    "Archive of superseded rows",
-  ]) {
-    const got = ids(body(marker));
-    assert.deepEqual(
-      got,
-      ["live"],
-      `a row under the «${marker}» marker is still counted as current — a superseded snapshot ` +
-        `reads as today's status, which is the exact failure the cut exists to prevent: ${JSON.stringify(got)}`,
-    );
-  }
-
-  // The other half. A heading that merely mentions neither word must NOT cut, or the check
-  // silently stops seeing whole sections and reports a clean scorecard it never read.
-  const kept = ids(body("Notes"));
-  assert.deepEqual(
-    kept,
-    ["live", "buried"],
-    `an ordinary heading cut the document — sections after it vanish unnoticed: ${JSON.stringify(kept)}`,
-  );
-}