chore(reviewer): add epic verification coverage

.security-review/suppressions.json

@@ -1,25 +1,5 @@
 {
   "suppressions": [
-    {
-      "id": "AUTHZ-CLEAN-AUTOAPPROVE-INJECTION-001",
-      "title": "Clean-review auto-approve derives APPROVE authority from a model-controlled signal on the untrusted auto-review path (prompt-injection-mintable APPROVE)",
-      "file": "agent/tools/publish_review.py",
-      "severity": "high",
-      "status": "confirmed",
-      "suppression_justification": "ACCEPTED (Adam, 2026-07-21) as a knowingly-deferred risk merged into the dev integration branch via admin merge in PR #217. /sh-security-review returned BLOCK: moving `approve` authorization from the deterministic verdict_requested flag to open_findings_count==0 lets a prompt-injected auto-review (which never sets verdict_requested) land a real bot APPROVE on an attacker-controlled external/fork PR, potentially satisfying branch protection. This reintroduces the hole PR #214 closed. Merged to dev only (NOT main/prod). Compensating controls: (a) confirm dev does not auto-deploy to sh-openswe; (b) on sensitive repos, esp. payments-dashboard, configure rulesets so the seahaven-openswe[bot] APPROVE does not by itself satisfy required approvals. Tracked in issue #218 with the full redesign checklist. REVISIT TRIGGER: MUST be resolved before this change promotes from dev to main/prod, and immediately if dev is found to auto-deploy. Verified HIGH by the sh-security-review detector fan-out (6 independent detectors converged).",
-      "owner": "adam@seahavenind.com",
-      "added": "2026-07-21"
-    },
-    {
-      "id": "AUTHZ-CLEAN-AUTOAPPROVE-THREADLAUNDER-002",
-      "title": "Clean-review auto-approve gate reads reconcile-derived finding status, so a PR author can launder a dirty PR to clean via GitHub thread resolve/outdate",
-      "file": "agent/tools/publish_review.py",
-      "severity": "high",
-      "status": "confirmed",
-      "suppression_justification": "ACCEPTED (Adam, 2026-07-21), deferred with AUTHZ-CLEAN-AUTOAPPROVE-INJECTION-001 in PR #217 (admin-merged to dev). The open_findings_count gate reads finding status after reconcile_findings_with_review_threads, which flips open->resolved when the GitHub thread is is_resolved/is_outdated — both author-controllable ('Resolve conversation' or a trivial hunk-outdating commit) without fixing the defect, yielding an auto-approve on an unfixed PR. Same compensating controls and revisit trigger as ...-001. Tracked in issue #218 (redesign: compute the gate from the reviewer's own authoritative finding state, not author-influenced thread state). Verified HIGH by /sh-security-review.",
-      "owner": "adam@seahavenind.com",
-      "added": "2026-07-21"
-    },
     {
       "id": "AUTHZ-SLACK-BOT-DEFAULT-001",
       "title": "Slack entrypoint lacks a per-user repo-access check; default-bot PR authoring removes the implicit per-user repo boundary",

CLAUDE.md

@@ -82,7 +82,7 @@ The system prompt instructs the agent to call a tool every turn, and `ensure_no_
 
 Other middleware exists in `agent/middleware/` (`ExcludeToolsMiddleware`) but isn't wired into the default agent. The reviewer uses a leaner stack (see `reviewer.py:get_reviewer_agent` for the authoritative order), including `SanitizeToolInputsMiddleware`, `ModelCallLimitMiddleware`, `ToolErrorMiddleware`, `PullRequestVerdictGuardMiddleware`, `SlackAssistantStatusMiddleware`, and `settle_review_check_on_exit`.
 
-**Verdict gating (Sea Haven fork):** `PullRequestVerdictGuardMiddleware` (`agent/middleware/pr_verdict_guard.py`) is wired into BOTH graphs (`server.py:get_agent` after `PullRequestCreationGuardMiddleware`; `reviewer.py:get_reviewer_agent` after `ToolErrorMiddleware`). It blocks shell-path review verdicts (`gh pr review --approve/-a/--request-changes/-r`, `gh api`/`curl` posting `event=APPROVE|REQUEST_CHANGES` to `/pulls/N/reviews`); comment reviews and reads pass through. Verdicts go exclusively through `publish_review(verdict=...)`: `request_changes` is honored only when the run's `configurable["verdict_requested"] is True` — set solely by the explicit-mention dispatch path (`trigger_pr_review_from_ref(request_verdict=True)` → `_build_reviewer_configurable`); auto-review dispatches never set it. `approve` is additionally honored on non-verdict-requested runs when the review is clean (zero open findings — the clean-review auto-approve); with open findings it downgrades to a comment (`verdict_ignored_reason="approve_with_open_findings"`). The tool layer also downgrades self-reviews (PR author in `INTERNAL_BOT_LOGINS`) to comment reviews and best-effort dismisses a recorded stale APPROVE when a later publish surfaces new findings.
+**Verdict gating (Sea Haven fork):** Verdict authority is set by dispatch, not inferred by the reviewer model. `verdict_requested=True` is reserved for explicit human paths, including a direct `@openswe review` command and the main agent's `request_pr_review` tool when the user explicitly asks for a verdict. `verdict_authorized=True` is reserved for automatic paths after deterministic policy and trust checks. Automatic verdicts default off at both the team and per-repository levels; either opt-in enables policy evaluation, but only a non-fork PR whose author is an internal bot or an active member of the repository organization receives automatic authority. Automatic verdicts must match authoritative finding state: `approve` requires zero blocking findings and `request_changes` requires one or more. Explicit human verdict requests are not subject to that automatic consistency rule, but both paths still re-check the current PR head, GitHub's recorded review state, and self-review constraints. A self-review is downgraded to a comment review; its `Open SWE Review` check remains finding-aware and can still fail when blocking findings exist. `PullRequestVerdictGuardMiddleware` is wired into both the coding and reviewer graphs, and all agent shell verdict paths are blocked: `gh pr review` verdict flags plus direct `gh api`/`curl` `APPROVE` or `REQUEST_CHANGES` submissions. Comment reviews and reads remain allowed. Verdicts must go through `publish_review(verdict=...)`, which also best-effort dismisses a recorded stale approval when a later review surfaces new findings.
 
 There is intentionally no after-agent safety net that opens a PR for the agent. The agent itself is responsible for committing, pushing, opening/updating the draft PR, and replying in the source channel — all via `GH_TOKEN=dummy gh` and `slack_thread_reply` / `linear_comment`.
 

tests/e2e/e2e_env.py

@@ -85,6 +85,8 @@ OTHER_USER = {"login": TEST_USERS[1]["login"], "email": TEST_USERS[1]["email"]}
 for _k, _v in _DEFAULTS.items():
     os.environ.setdefault(_k, _v)
 
+os.environ["CONFIGURED_ADMINS"] = "alice,alice@example.com"
+
 for _d in (TMP, _GH_DIR, _WORK_DIR):
     _d.mkdir(parents=True, exist_ok=True)
 

tests/e2e/fakes.py

@@ -69,7 +69,10 @@ def slack_messages(channel: str) -> list[dict[str, Any]]:
 
 # --- GitHub ----------------------------------------------------------------
 PULLS: list[dict[str, Any]] = []
+CHECK_RUNS: list[dict[str, Any]] = []
+REVIEW_DISPATCHES: list[dict[str, Any]] = []
 _pr_seq = [0]
+_check_seq = [0]
 
 
 def _git(*args: str, cwd: Path | None = None) -> str:
@@ -151,6 +154,8 @@ def create_pull(
         "state": "open",
         "merged": False,
         "author": "open-swe[bot]",
+        "head_sha": f"head-{number:04d}",
+        "base_sha": f"base-{number:04d}",
         "files": files,
         "additions": sum(f["additions"] for f in files),
         "deletions": sum(f["deletions"] for f in files),
@@ -159,12 +164,56 @@ def create_pull(
     return pr
 
 
+def create_review_pull(owner: str, repo: str) -> dict[str, Any]:
+    pr = create_pull(
+        owner,
+        repo,
+        head="feature/review-me",
+        base=BASE_BRANCH,
+        title="Review command fixture",
+        body="A deterministic pull request for reviewer routing.",
+        draft=False,
+    )
+    pr["author"] = "alice"
+    return pr
+
+
 def find_pull(number: int) -> dict[str, Any] | None:
     return next((p for p in PULLS if p["number"] == number), None)
 
 
+def create_check_run(owner: str, repo: str, payload: dict[str, Any]) -> dict[str, Any]:
+    _check_seq[0] += 1
+    check = {
+        "id": _check_seq[0],
+        "owner": owner,
+        "repo": repo,
+        "name": payload.get("name"),
+        "head_sha": payload.get("head_sha"),
+        "status": payload.get("status"),
+        "conclusion": payload.get("conclusion"),
+        "details_url": payload.get("details_url"),
+        "output": payload.get("output", {}),
+    }
+    CHECK_RUNS.append(check)
+    return check
+
+
+def update_check_run(check_run_id: int, payload: dict[str, Any]) -> dict[str, Any] | None:
+    check = next((item for item in CHECK_RUNS if item["id"] == check_run_id), None)
+    if check is None:
+        return None
+    check.update(
+        {key: payload[key] for key in ("status", "conclusion", "output") if key in payload}
+    )
+    return check
+
+
 def reset() -> None:
     SLACK_MESSAGES.clear()
     PULLS.clear()
+    CHECK_RUNS.clear()
+    REVIEW_DISPATCHES.clear()
     _pr_seq[0] = 0
+    _check_seq[0] = 0
     seed_bare_remote()

tests/e2e/harness.py

@@ -60,8 +60,11 @@ _SLACK_USERS: dict[str, dict[str, str]] = {
 }
 
 from agent.api.app import app  # noqa: E402
+from agent.dashboard import routes as dashboard_routes  # noqa: E402
 from agent.dashboard.oauth import COOKIE_NAME, issue_session  # noqa: E402
+from agent.utils import github_checks  # noqa: E402
 from agent.utils.thread_ids import generate_thread_id_from_slack_thread  # noqa: E402
+from agent.webhooks import common as webhook_common  # noqa: E402
 
 GITHUB_WEBHOOK_SECRET = os.environ["GITHUB_WEBHOOK_SECRET"]
 SLACK_SIGNING_SECRET = os.environ["SLACK_SIGNING_SECRET"]
@@ -70,6 +73,86 @@ STATIC_DIR = Path(__file__).parent / "static"
 CURRENT_THREAD: dict[str, str | None] = {"channel": DEMO_CHANNEL, "thread_ts": None}
 
 fakes.seed_bare_remote()
+_real_dispatch_agent_run = webhook_common.dispatch_agent_run
+
+
+async def _fake_installation_token(*_args: object, **_kwargs: object) -> str:
+    return "dummy-installation-token"
+
+
+async def _fake_installation_token_with_expiry(
+    *_args: object, **_kwargs: object
+) -> tuple[str, None]:
+    return "dummy-installation-token", None
+
+
+async def _fake_fetch_pr_metadata(pr_ref: Any, *, token: str) -> dict[str, Any] | None:  # noqa: ARG001
+    pr = fakes.find_pull(pr_ref.number)
+    return _gh_pr_json(pr) if pr is not None else None
+
+
+async def _fake_reviewer_token(*_args: object, **_kwargs: object) -> tuple[str, None]:
+    return "dummy-installation-token", None
+
+
+async def _fake_reaction(*_args: object, **_kwargs: object) -> bool:
+    return True
+
+
+async def _fake_started_comment(*_args: object, **_kwargs: object) -> None:
+    return None
+
+
+async def _record_review_dispatch(
+    thread_id: str,
+    prompt: str,
+    configurable: dict[str, Any],
+    *,
+    source: str,
+    assistant_id: str = "agent",
+    **kwargs: object,
+) -> dict[str, str]:
+    if assistant_id != "reviewer":
+        return await _real_dispatch_agent_run(
+            thread_id,
+            prompt,
+            configurable,
+            source=source,
+            assistant_id=assistant_id,
+            **kwargs,
+        )
+    run_id = f"review-run-{len(fakes.REVIEW_DISPATCHES) + 1}"
+    fakes.REVIEW_DISPATCHES.append(
+        {
+            "thread_id": thread_id,
+            "prompt": prompt,
+            "configurable": configurable,
+            "source": source,
+            "assistant_id": assistant_id,
+            "run_id": run_id,
+        }
+    )
+    return {"run_id": run_id}
+
+
+async def _fake_installations_and_repos(
+    _login: str,
+) -> tuple[list[dict[str, Any]], list[dict[str, Any]]]:
+    return (
+        [{"id": 1, "account": {"login": e2e_env.OWNER, "type": "Organization"}}],
+        [{"full_name": f"{e2e_env.OWNER}/{e2e_env.REPO}", "private": False}],
+    )
+
+
+webhook_common.get_github_app_installation_token = _fake_installation_token
+webhook_common.get_github_app_installation_token_with_expiry = _fake_installation_token_with_expiry
+webhook_common.fetch_github_pr_metadata = _fake_fetch_pr_metadata
+webhook_common._reviewer_token_for_repo = _fake_reviewer_token
+webhook_common.react_to_github_comment = _fake_reaction
+webhook_common.post_review_started_comment = _fake_started_comment
+webhook_common.dispatch_agent_run = _record_review_dispatch
+dashboard_routes._fetch_user_installations_and_repos = _fake_installations_and_repos
+github_checks._GITHUB_API_BASE = e2e_env.FAKE_GITHUB_API
 
 
 # --- control + Slack compose (the test driver) -----------------------------
@@ -80,6 +163,53 @@ async def control_reset() -> JSONResponse:
     return JSONResponse({"ok": True})
 
 
+@app.post("/control/review-pr")
+async def control_review_pr() -> JSONResponse:
+    fakes.reset()
+    pr = fakes.create_review_pull(e2e_env.OWNER, e2e_env.REPO)
+    return JSONResponse(_gh_pr_json(pr))
+
+
+@app.get("/control/review-dispatches")
+async def control_review_dispatches() -> JSONResponse:
+    return JSONResponse(fakes.REVIEW_DISPATCHES)
+
+
+@app.get("/control/check-runs")
+async def control_check_runs() -> JSONResponse:
+    return JSONResponse(fakes.CHECK_RUNS)
+
+
+@app.post("/control/review-check/evaluate")
+async def control_review_check_evaluate(request: Request) -> JSONResponse:
+    body = await request.json()
+    outcome = body.get("outcome")
+    if not isinstance(outcome, dict):
+        raise HTTPException(400, "outcome must be an object")
+    head_sha = str(body.get("head_sha") or f"evaluation-{len(fakes.CHECK_RUNS) + 1}")
+    check_run_id = await github_checks.create_review_check_run(
+        owner=e2e_env.OWNER,
+        repo=e2e_env.REPO,
+        head_sha=head_sha,
+        token="dummy-installation-token",
+    )
+    if check_run_id is None:
+        raise HTTPException(500, "failed to create check run")
+    conclusion, title, summary = github_checks.review_check_conclusion(outcome)
+    completed = await github_checks.complete_review_check_run(
+        owner=e2e_env.OWNER,
+        repo=e2e_env.REPO,
+        check_run_id=check_run_id,
+        token="dummy-installation-token",
+        conclusion=conclusion,
+        title=title,
+        summary=summary,
+    )
+    if not completed:
+        raise HTTPException(500, "failed to complete check run")
+    return JSONResponse({"id": check_run_id, "conclusion": conclusion, "title": title})
+
+
 @app.get("/control/state")
 async def control_state() -> JSONResponse:
     return JSONResponse(
@@ -313,6 +443,16 @@ async def ui_agents_plan(thread_id: str) -> FileResponse:  # noqa: ARG001
     return _ui_file("_shell.html")
 
 
+@app.get("/review", response_class=HTMLResponse)
+async def ui_review() -> FileResponse:
+    return _ui_file("_shell.html")
+
+
+@app.get("/review/repositories/{owner}", response_class=HTMLResponse)
+async def ui_review_repositories(owner: str) -> FileResponse:  # noqa: ARG001
+    return _ui_file("_shell.html")
+
+
 @app.get("/login", response_class=HTMLResponse)
 async def ui_login() -> FileResponse:
     return _ui_file("_shell.html")
@@ -406,6 +546,8 @@ async def mock_github_pr(owner: str, repo: str, number: int) -> HTMLResponse:  #
 
 # --- fake GitHub REST API (open_pull_request hits this) --------------------
 def _gh_pr_json(pr: dict[str, Any]) -> dict[str, Any]:
+    full_name = f"{pr['owner']}/{pr['repo']}"
+    repo = {"id": 1, "full_name": full_name, "private": False}
     return {
         "number": pr["number"],
         "html_url": _pr_html_url(pr),
@@ -415,8 +557,8 @@ def _gh_pr_json(pr: dict[str, Any]) -> dict[str, Any]:
         "title": pr["title"],
         "body": pr["body"],
         "user": {"login": pr["author"]},
-        "head": {"ref": pr["head"]},
-        "base": {"ref": pr["base"]},
+        "head": {"ref": pr["head"], "sha": pr["head_sha"], "repo": repo},
+        "base": {"ref": pr["base"], "sha": pr["base_sha"], "repo": repo},
         "additions": pr["additions"],
         "deletions": pr["deletions"],
         "changed_files": len(pr["files"]),
@@ -463,6 +605,26 @@ async def gh_get_pull(owner: str, repo: str, number: int) -> JSONResponse:  # no
     return JSONResponse(_gh_pr_json(pr))
 
 
+@app.post("/fake-gh/repos/{owner}/{repo}/check-runs")
+async def gh_create_check_run(owner: str, repo: str, request: Request) -> JSONResponse:
+    payload = await request.json()
+    return JSONResponse(fakes.create_check_run(owner, repo, payload), status_code=201)
+
+
+@app.patch("/fake-gh/repos/{owner}/{repo}/check-runs/{check_run_id}")
+async def gh_update_check_run(
+    owner: str,
+    repo: str,
+    check_run_id: int,
+    request: Request,  # noqa: ARG001
+) -> JSONResponse:
+    payload = await request.json()
+    check = fakes.update_check_run(check_run_id, payload)
+    if check is None:
+        return JSONResponse({"message": "Not Found"}, status_code=404)
+    return JSONResponse(check)
+
+
 # --- fake Slack API (real slack code hits this) ----------------------------
 def _ok(extra: dict[str, Any] | None = None) -> JSONResponse:
     return JSONResponse({"ok": True, **(extra or {})})

tests/e2e/tests/dashboard.spec.ts

@@ -56,6 +56,85 @@ async function expectTranscriptVisible(page: Page) {
   }).toPass({ timeout: 60000 });
 }
 
+test.describe("Review verdict settings (real dashboard UI and API)", () => {
+  test("admin controls persist team and per-repository auto-verdict policy", async ({
+    page,
+    baseURL,
+  }) => {
+    await loginAs(page, SAME_USER);
+    const mutationHeaders = { origin: baseURL ?? "" };
+
+    const resetTeam = await page.request.put("/dashboard/api/team-settings", {
+      data: { auto_verdict: false },
+      headers: mutationHeaders,
+    });
+    expect(resetTeam.ok(), await resetTeam.text()).toBeTruthy();
+    const resetRepo = await page.request.put(
+      "/dashboard/api/auto-verdict-repos",
+      {
+        data: { full_name: "fakeorg/demo", enabled: false },
+        headers: mutationHeaders,
+      },
+    );
+    expect(resetRepo.ok()).toBeTruthy();
+
+    await page.goto("/review");
+    await expect(
+      page.getByRole("heading", { name: "Open SWE Review" }),
+    ).toBeVisible();
+    const teamToggle = page.getByRole("switch").first();
+    await expect(teamToggle).not.toBeChecked();
+    const teamSaved = page.waitForResponse(
+      (response) =>
+        response.url().endsWith("/dashboard/api/team-settings") &&
+        response.request().method() === "PUT",
+    );
+    await teamToggle.click();
+    expect((await teamSaved).ok()).toBeTruthy();
+    await expect(teamToggle).toBeChecked();
+
+    const teamContract = await (
+      await page.request.get("/dashboard/api/team-settings")
+    ).json();
+    expect(teamContract.auto_verdict).toBe(true);
+
+    await page.goto("/review/repositories/fakeorg");
+    const repoToggle = page.getByRole("switch", {
+      name: "Allow automatic verdicts for fakeorg/demo",
+    });
+    await expect(repoToggle).not.toBeChecked();
+    const repoSaved = page.waitForResponse(
+      (response) =>
+        response.url().endsWith("/dashboard/api/auto-verdict-repos") &&
+        response.request().method() === "PUT",
+    );
+    await repoToggle.click();
+    expect((await repoSaved).ok()).toBeTruthy();
+    await expect(repoToggle).toBeChecked();
+
+    const repoContract = await (
+      await page.request.get("/dashboard/api/auto-verdict-repos")
+    ).json();
+    expect(repoContract).toEqual({ repos: ["fakeorg/demo"] });
+
+    await loginAs(page, OTHER_USER);
+    const forbidden = await page.request.put(
+      "/dashboard/api/auto-verdict-repos",
+      {
+        data: { full_name: "fakeorg/demo", enabled: false },
+        headers: mutationHeaders,
+      },
+    );
+    expect(forbidden.status()).toBe(403);
+    await page.reload();
+    await expect(
+      page.getByRole("switch", {
+        name: "Allow automatic verdicts for fakeorg/demo",
+      }),
+    ).toBeDisabled();
+  });
+});
+
 test.describe("Slack → web handoff (real dashboard UI)", () => {
   test("the SAME user continues the conversation in the web app", async ({
     page,

tests/e2e/tests/reviewer_verification.spec.ts

@@ -0,0 +1,145 @@
+import { createHmac } from "node:crypto";
+
+import { expect, test } from "@playwright/test";
+
+const WEBHOOK_SECRET = "test-github-secret";
+
+test.describe("Reviewer verification contracts", () => {
+  test("direct @openswe review routes to the reviewer without user mapping", async ({
+    page,
+  }) => {
+    const seeded = await page.request.post("/control/review-pr");
+    expect(seeded.ok()).toBeTruthy();
+    const pr = await seeded.json();
+    const payload = {
+      action: "created",
+      repository: {
+        id: 1,
+        name: "demo",
+        full_name: "fakeorg/demo",
+        private: true,
+        owner: { login: "fakeorg" },
+      },
+      issue: {
+        number: pr.number,
+        html_url: pr.html_url,
+        pull_request: { html_url: pr.html_url },
+      },
+      comment: {
+        id: 24601,
+        node_id: "IC_kwDO_e2e",
+        body: "@openswe review: focus on authorization boundaries",
+        user: { login: "unmapped-reviewer" },
+      },
+      sender: { id: 9001, login: "unmapped-reviewer" },
+    };
+    const raw = JSON.stringify(payload);
+    const signature = `sha256=${createHmac("sha256", WEBHOOK_SECRET)
+      .update(raw)
+      .digest("hex")}`;
+
+    const webhook = await page.request.post("/webhooks/github", {
+      data: raw,
+      headers: {
+        "Content-Type": "application/json",
+        "X-GitHub-Event": "issue_comment",
+        "X-Hub-Signature-256": signature,
+      },
+    });
+    expect(webhook.ok()).toBeTruthy();
+    expect(await webhook.json()).toEqual({
+      status: "accepted",
+      message: "Processing on-demand PR review",
+    });
+
+    await expect
+      .poll(async () => {
+        const response = await page.request.get("/control/review-dispatches");
+        return (await response.json()).length;
+      })
+      .toBe(1);
+    const dispatches = await (
+      await page.request.get("/control/review-dispatches")
+    ).json();
+    expect(dispatches).toHaveLength(1);
+    expect(dispatches[0].assistant_id).toBe("reviewer");
+    expect(dispatches[0].source).toBe("github_comment");
+    expect(dispatches[0].configurable).toMatchObject({
+      github_login: "unmapped-reviewer",
+      github_user_id: 9001,
+      review_requested: true,
+      verdict_requested: true,
+      repo: { owner: "fakeorg", name: "demo" },
+      pr_number: pr.number,
+    });
+    expect(dispatches[0].configurable).not.toHaveProperty("email");
+    expect(dispatches[0].prompt).toContain(
+      "focus on authorization boundaries",
+    );
+
+  });
+
+  test("review outcomes settle checks as success, failure, or neutral", async ({
+    page,
+  }) => {
+    await page.request.post("/control/reset");
+    const cases = [
+      {
+        outcome: {
+          verdict_submitted: true,
+          verdict_event: "APPROVE",
+          blocking_finding_count: 0,
+        },
+        conclusion: "success",
+        title: "Review approved",
+      },
+      {
+        outcome: {
+          verdict_submitted: true,
+          verdict_event: "REQUEST_CHANGES",
+          blocking_finding_count: 1,
+        },
+        conclusion: "failure",
+        title: "Changes requested",
+      },
+      {
+        outcome: {
+          verdict_submitted: false,
+          verdict_ignored_reason: "verdict_not_requested",
+          blocking_finding_count: 0,
+        },
+        conclusion: "neutral",
+        title: "Verdict withheld",
+      },
+    ];
+
+    for (const [index, item] of cases.entries()) {
+      const response = await page.request.post(
+        "/control/review-check/evaluate",
+        {
+          data: {
+            head_sha: `tri-state-${index + 1}`,
+            outcome: item.outcome,
+          },
+        },
+      );
+      expect(response.ok()).toBeTruthy();
+      expect(await response.json()).toMatchObject({
+        conclusion: item.conclusion,
+        title: item.title,
+      });
+    }
+
+    const checks = await (
+      await page.request.get("/control/check-runs")
+    ).json();
+    expect(checks.map((check: { conclusion: string }) => check.conclusion)).toEqual([
+      "success",
+      "failure",
+      "neutral",
+    ]);
+    expect(
+      checks.map((check: { status: string }) => check.status),
+    ).toEqual(["completed", "completed", "completed"]);
+  });
+});