Single-reviewer approval gate and clearer awaiting-review status

.github/risk-tiers.yml

@@ -87,7 +87,7 @@ high:
     # Applied by the Contributor Reputation Check writer.
     - "needs-review:HIGH"
   approvals:
-    required: 2
+    required: 1
     require_core: true
 
 medium:

docs/maintainers/submission-gate.md

@@ -131,7 +131,7 @@ The status comment includes a collapsed "Why this tier" list with the reasons th
 |---|---|
 | `merge-risk:low` | 1 approval from an owner of the changed resource: its domain pool, or the core pools for files outside every domain |
 | `merge-risk:medium` | 1 approval from a domain reviewer for the area touched |
-| `merge-risk:high` | 2 approvals, including a core maintainer |
+| `merge-risk:high` | 1 approval from a core maintainer |
 
 How approvals are counted:
 
@@ -153,7 +153,9 @@ The core pool is `core-maintainers`. Its members also satisfy the domain and res
 If routing isn't staffed yet, the gate falls back instead of blocking:
 
 - **`.github/review-routing.yml` missing, or the matching domain pools empty:** any approver with write access satisfies the domain or resource-owner requirement. If the file exists but can't be parsed, or `pools` isn't a mapping, the gate fails instead of falling back, so a broken edit can't relax approvals.
-- **`core-maintainers` pool empty:** a high-risk PR needs one of its two approvals from a user with `admin` or `maintain` permission.
+- **`core-maintainers` pool empty:** a high-risk PR needs its approval from a user with `admin` or `maintain` permission.
+
+The `submission-gate` check reports `action_required` ("Action required") rather than `failure` when the only thing blocking a PR is a missing approval, so it reads as "waiting for a reviewer" instead of a broken workflow. The check summary lists exactly what is missing. It still blocks merging like any non-passing required check. Failed checks and requested changes still report `failure`.
 
 The status comment shows a note whenever a fallback is in effect.
 
@@ -223,7 +225,7 @@ Commands use the same reader/writer split as the gate, so they work on PRs from
 - **Only trusted code runs.** No workflow here executes PR code with a write token. The gate loads its logic from the base commit, except in the bootstrap case described above, which still has only a read-only token. The writer and command workflows check out the default branch and install dependencies with `npm ci --ignore-scripts`.
 - **Contributor reputation artifact.** It is used only as a raise-only signal. It must match schema `contributor-check-result/v1` and the PR head SHA. If the `pr-check` job succeeded but the artifact is missing or unreadable, the gate records an infrastructure failure and blocks until a later sweep reads it, so a lost `HIGH` signal can't lower the tier.
 - **Untrusted text.** Comment bodies are never interpolated into scripts: the reader passes them through environment variables, and the writer re-reads them from the API. Reviewer logins, check names, and details are sanitized before they go into markdown.
-- **Tampering with other checks.** A PR can also edit the validation workflows the gate waits for. Any such edit touches `.github/**`, which makes the PR `merge-risk:high`: 2 approvals, including a core maintainer. The writer computes that tier from default-branch code.
+- **Tampering with other checks.** A PR can also edit the validation workflows the gate waits for. Any such edit touches `.github/**`, which makes the PR `merge-risk:high`: 1 approval from a core maintainer. The writer computes that tier from default-branch code.
 - **Check source.** Check runs created with `GITHUB_TOKEN` belong to the GitHub Actions app, the same app as `pull_request` jobs. Requiring `submission-gate` from GitHub Actions in the ruleset blocks other apps and commit statuses; impostor detection covers jobs in the same app.
 - **Events.** Check runs created with `GITHUB_TOKEN` don't trigger `check_run` workflows. Automation that reacts to the gate should listen for `workflow_run` on **Submission Gate Writer**.
 - **Rate limits.** The hourly writer sweep refreshes at most 60 open PRs updated in the last 30 days, so a large backlog stays within API rate limits.

eng/submission-gate.mjs

@@ -1097,6 +1097,22 @@ export async function findImpostorGateChecks(github, { owner, repo, headSha }) {
   return runs.filter((run) => run.external_id !== GATE_CHECK_EXTERNAL_ID);
 }
 
+/**
+ * A PR that is only waiting for a review is not broken: report `action_required`, which
+ * still blocks the required check but reads as "needs a reviewer" rather than a failure.
+ */
+export function gateConclusion(evaluation) {
+  if (evaluation.passed) return "success";
+  const { automation, approvals } = evaluation;
+  const onlyAwaitingReview =
+    automation.contributionFailures.length === 0 &&
+    automation.infrastructureFailures.length === 0 &&
+    automation.pending.length === 0 &&
+    (approvals.changesRequestedBy ?? []).length === 0 &&
+    !approvals.satisfied;
+  return onlyAwaitingReview ? "action_required" : "failure";
+}
+
 /**
  * Publish the required `submission-gate` check on the PR head commit. Only the trusted
  * writer calls this, so the check can't be satisfied by editing a PR-controlled workflow.
@@ -1114,7 +1130,7 @@ export async function publishGateCheck(github, { owner, repo, evaluation, detail
   };
   const fields = pending
     ? { status: "in_progress", output }
-    : { status: "completed", conclusion: evaluation.passed ? "success" : "failure", completed_at: new Date().toISOString(), output };
+    : { status: "completed", conclusion: gateConclusion(evaluation), completed_at: new Date().toISOString(), output };
   if (detailsUrl) fields.details_url = detailsUrl;
 
   const runs = await listGateCheckRuns(github, { owner, repo, headSha: evaluation.headSha });

eng/submission-gate.test.mjs

@@ -11,6 +11,7 @@ import {
   classifyRisk,
   computeState,
   evaluateApprovals,
+  gateConclusion,
   evaluateCheck,
   evaluateSubmission,
   globToRegExp,
@@ -274,17 +275,25 @@ test("medium tier requires a domain reviewer when the pool is staffed", () => {
   assert.equal(noRouting.satisfied, true);
 });
 
-test("high tier requires two approvals including a core maintainer", () => {
+test("high tier requires one approval from a core maintainer", () => {
   const permissions = new Map([["alice", "write"], ["bob", "write"], ["corea", "write"], ["admin1", "admin"]]);
   const files = [file(".github/workflows/x.yml")];
   const base = { tier: "high", tiers, author: "author", permissions, routing, files };
-  assert.equal(evaluateApprovals({ ...base, reviews: [review("corea", "APPROVED")] }).satisfied, false);
+  assert.equal(evaluateApprovals({ ...base, reviews: [review("corea", "APPROVED")] }).satisfied, true);
   assert.equal(evaluateApprovals({ ...base, reviews: [review("alice", "APPROVED"), review("bob", "APPROVED")] }).satisfied, false);
-  assert.equal(evaluateApprovals({ ...base, reviews: [review("alice", "APPROVED"), review("corea", "APPROVED")] }).satisfied, true);
 
   const fallback = { ...base, routing: null };
-  assert.equal(evaluateApprovals({ ...fallback, reviews: [review("alice", "APPROVED"), review("bob", "APPROVED")] }).satisfied, false);
-  assert.equal(evaluateApprovals({ ...fallback, reviews: [review("alice", "APPROVED"), review("admin1", "APPROVED")] }).satisfied, true);
+  assert.equal(evaluateApprovals({ ...fallback, reviews: [review("alice", "APPROVED")] }).satisfied, false);
+  assert.equal(evaluateApprovals({ ...fallback, reviews: [review("admin1", "APPROVED")] }).satisfied, true);
+});
+
+test("gateConclusion reports a PR that only awaits review as action_required", () => {
+  const clean = { contributionFailures: [], infrastructureFailures: [], pending: [] };
+  const waiting = { satisfied: false, changesRequestedBy: [] };
+  assert.equal(gateConclusion({ passed: true, automation: clean, approvals: { satisfied: true, changesRequestedBy: [] } }), "success");
+  assert.equal(gateConclusion({ passed: false, automation: clean, approvals: waiting }), "action_required");
+  assert.equal(gateConclusion({ passed: false, automation: clean, approvals: { ...waiting, changesRequestedBy: ["bob"] } }), "failure");
+  assert.equal(gateConclusion({ passed: false, automation: { ...clean, contributionFailures: [{}] }, approvals: waiting }), "failure");
 });
 
 // --- checks -------------------------------------------------------------------
@@ -904,7 +913,7 @@ test("publishGateCheck corrects copies of the writer's check that claim success
   await publishGateCheck(github, { owner: "github", repo: "awesome-copilot", evaluation, log: (line) => logs.push(line) });
   const updates = github.calls.filter((call) => call.name === "checks.update");
   assert.deepEqual(updates.map((call) => call.params.check_run_id), [11]);
-  assert.equal(updates[0].params.conclusion, "failure");
+  assert.equal(updates[0].params.conclusion, "action_required");
   assert.ok(logs.some((line) => /Correcting 1/.test(line)));
 });