Files
PaperClipAI/tests/runner-e2e/chat-catalog-domains/runtime-post-stop.ts
T
DottaandClaude Opus 5.5 5f34f777ec fix(evals): grade 16 chat cases by saved state instead of reply keywords (#20600)
## Thinking Path

> - Paperclip is the open source app people use to manage AI agents for
work
> - The chat-feature eval catalog checks that a chat agent does the work
a person asks for
> - In the latest full-cohort run on master, 16 cases failed although
the saved work was correct
> - Their graders matched words, one literal value, or one API route in
free-form model output
> - A grader that fails correct work hides real failures in the noise,
and it can also pass wrong work that uses the expected word
> - This pull request grades those cases by saved state, audit records,
setup literals and rendered page structure, and keeps only narrow checks
for false claims
> - The benefit is that these cases fail only when the product or agent
does the wrong thing

## Linked Issues or Issue Description

Refs #18971 (related: company-secret question handoffs)

**What happened**

The full-cohort chat-eval run on master failed 16 cases where the
retained evidence shows correct behaviour. Examples:
`companies.rename-company` accepted only `company.updated`, but the
supported branding route logs `company.branding_updated`.
`recovery.post-stop-message` required a signup button label that matched
a regex, and rejected a labelled email form with the button "Start
writing free". `secrets.secret-proposal` flagged "the value is delivered
when Morgan next runs" as a false delivery claim. Other graders required
words such as "uncommitted", "telemetry", "administrator" or the app
name that the question did not ask for.

**Expected behavior**

A grader accepts every supported route and wording that achieves the
outcome. It fails when saved state is wrong, when a requested fact is
wrong, or when the reply makes a claim that saved state contradicts.

**Steps to reproduce**

1. Load the retained observation for one of the cases below from the
full-cohort evidence.
2. Grade it with the old oracle. The keyword or route check fails
although the saved state is correct.
3. Grade it with the new oracle. It passes. A broken variant (wrong
saved state or a contradicting claim) still fails.

## What Changed

- `activity.audit-change`: attribution is graded by structure. The reply
names the recorded editor, names no other company agent (new setup
literal `otherActors`), and states only real transitions. Sentence
grammar is no longer matched.
- `app-permissions.ask-before-action`: each continuation reply must
exist and must link no app record other than the reviewed connection. It
no longer must name the app.
- `budgets-costs.spend-breakdown`: the reply must point at the record
with unresolved usage (task link, identifier or run id) and must not
give it a zero or free amount. The words "subscription" and "incomplete"
are no longer required.
- `cases.browse-cases`: an excluded case may be mentioned in prose. It
must not be linked or listed as a result row.
- `cases.handoff-same-task`: handoff instructions carry findings forward
when they point at the saved `findings` document, reproduce it, or
restate it. The premise (a primary Morgan run performs the handoff) is
unchanged.
- `cases.update-case-facts`: a saved browser value may add clauses that
only rule out other platforms ("desktop Safari unaffected").
- `companies.agent-observability`: the reply no longer must name
Telemetry and the run log. It must not claim that they are exported.
- `companies.rename-company`: both supported name routes
(`company.updated`, `company.branding_updated`) are accepted.
- `connection-setup.choose-provider`: the answer must exist and must not
claim a connected account. The unavailable method no longer must be
named; the setup path stays graded by clicking the card or link.
- `goals.align-existing-work`: a question card from the answering run
counts as the phase-two response. Refusal words and "unchanged" words
are no longer required; false change claims still fail.
- `recovery.post-stop-message` (and `steering.interrupt-with-direction`,
which shares the helper): a visible form with a labelled field and a
submit control is a signup action, whatever the button says.
- `routines.pause-owned`: the "paused"/"running" words are no longer
required. The saved checkpoint proves both; a cancellation claim still
fails.
- `secrets.secret-proposal` (and `secrets.configure-company-secret`,
which shares the helper): a sentence that says when delivery will happen
is not a delivery claim.
- `skills.explicit-restriction`: the reply no longer must say
"administrator". Success claims and invented tool limitations still
fail.
- `workspaces.inspect-git`: the reply no longer must say "uncommitted".
Unchanged HEAD and status prove it; commit or push claims still fail.
- Every affected `oracleVersion` is bumped in code and in
`chat-catalog-descriptors.json`. No scenario YAML changed.

## Verification

- `pnpm exec vitest run --config tests/runner-e2e/vitest.config.ts
--testTimeout=180000 tests/runner-e2e/chat-catalog-domains
tests/runner-e2e/chat-catalog-contract.test.ts
tests/runner-e2e/chat-catalog-consistency.test.ts`: all pass.
- `PAPERCLIP_CHAT_EVAL_RENDERED_ARTIFACT_CANARY=1` rendered-artifact
canary with Chromium: pass, including new form cases.
- `node --test
tests/runner-e2e/chat-full-catalog-testops-manifest.test.mjs`: pass.
`node tests/runner-e2e/source-sdk-public-inputs.mjs <HEAD> --check`:
unchanged pins.
- `npx tsc --noEmit -p tests/runner-e2e`: no errors in changed files.
- Offline calibration on retained evidence, for each case: the retained
correct evidence passes, earlier passing evidence passes where it
exists, and a broken variant fails. `spend-breakdown` retained evidence
still fails, because its reply shows unresolved usage as `$0.00`.
`handoff-same-task` retained evidence still fails on the kept
primary-run premise.
- Live paid runs: see the comment below.

## Risks

- Low risk. The change is limited to eval graders and their unit tests.
Product code does not change.
- Each grader is more accepting of wording. Saved-state checks and the
narrow false-claim checks remain, so a wrong outcome still fails.
- `secrets.configure-company-secret` still fails its retained evidence.
Its agent asked a two-question card, and the act supports only
one-question cards. This pull request changes it only by the shared
delivery-claim helper.

## Model Used

- Anthropic Claude Opus 5.5 (`claude-opus-5-5`), tool use and code
execution in Claude Code.

## Checklist

- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [ ] All Paperclip CI gates are green
- [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 13:37:16 -06:00

139 lines
19 KiB
TypeScript

import { expect, type Page } from "@playwright/test";
import { writeFile } from "node:fs/promises";
import type { ChatCatalogContext, NativeStopBoundaryReceipt } from "../chat-catalog-contract.js";
import type { IssueComment } from "../../../packages/shared/src/types/issue.js";
import { pollUntil } from "../api.js";
import { sendChatMessage } from "../chat-flow.js";
import { collectRunEvents } from "../run-observations.js";
import { readRegisteredArtifacts } from "../registered-artifact.js";
import { checks } from "./internal-support.js";
import { openStandingChat } from "./identity-support.js";
import { prepareRuntimeWorkerGate, readRuntimeTaskGate } from "./runtime-busy-worker.js";
import { suspendOwnedProviderForCleanup, releaseOwnedCleanupProvider, ownedCleanupProviderHasExited, ownedCleanupProviderIsHeld } from "./runtime-provider-cleanup.js";
import type { NativeFaultRun } from "./runtime-native-fault.js";
import { hasSimplerWebsite, renderRuntimeArtifact, simplerWebsiteChecks } from "./runtime-rendered-artifact.js";
import { observeRuntime, runtimeDefinition, runtimeObservation, type RuntimeAgent } from "./runtime-support.js";
interface StopIntent { schema?: string; intentId?: string; intentAuditId?: string; companyId?: string; runId?: string; issueId?: string; scope?: string; dispatchState?: string; dispatched?: boolean; recordedAt?: string; acknowledgedAt?: string; acknowledgementAuditId?: string }
interface StopHandle { page: Page; completed?: Promise<number> }
const handles = new WeakMap<ChatCatalogContext, StopHandle>();
export function sameSourcePostStopChronology(input: { boundary: NativeStopBoundaryReceipt; old: NativeFaultRun; next: NativeFaultRun; human: Pick<IssueComment, "id" | "body" | "issueId" | "companyId" | "authorAgentId" | "authorUserId" | "createdAt">; query: string; companyId: string; agentId: string; humanId: string; prompt: string; wakeCommentIds: string[]; heldReleasedAt: string }) {
const { old, next, human, boundary: b } = input, intent = old.resultJson?.nativeCancellation as StopIntent | undefined;
const before = Date.parse(b.intentRecordedAt), sent = Date.parse(String(human.createdAt)), ack = Date.parse(intent?.acknowledgedAt ?? ""), started = Date.parse(String(next.startedAt)), ended = Date.parse(String(old.finishedAt)), released = Date.parse(input.heldReleasedAt);
return old.id === b.runId && old.nativeIssueId === b.issueId && old.companyId === input.companyId && old.agentId === input.agentId && old.runtimeMode === "native" && old.status === "cancelled"
&& intent?.schema === "paperclip.native-cancellation.v1" && intent.intentId === b.intentId && intent.issueId === b.issueId && intent.companyId === input.companyId && intent.runId === old.id && intent.scope === "run"
&& !!intent.intentAuditId && !!intent.acknowledgementAuditId && intent.recordedAt === b.intentRecordedAt && intent.dispatchState === "acknowledged" && intent.dispatched === true
&& human.issueId === b.issueId && human.companyId === input.companyId && !human.authorAgentId && human.authorUserId === input.humanId && human.body === input.query
&& before <= sent && sent < released && Number.isFinite(ack) && next.id !== old.id && next.nativeIssueId === b.issueId && next.companyId === input.companyId && next.agentId === input.agentId && next.runtimeMode === "native" && next.status === "succeeded"
&& started >= ack && started >= ended && started >= released && input.wakeCommentIds.length === 1 && input.wakeCommentIds[0] === human.id && input.prompt.includes(human.id) && input.prompt.includes(input.query);
}
export async function releasePostStopGateAfterCleanup(old: Pick<NativeFaultRun, "status" | "resultJson">, providerExited: boolean, release: () => Promise<void>) {
const intent = old.resultJson?.nativeCancellation as StopIntent | undefined;
if (old.status !== "cancelled" || intent?.dispatchState !== "acknowledged" || intent.dispatched !== true || !intent.acknowledgedAt || !intent.acknowledgementAuditId || !providerExited)
throw new Error("The old native cleanup must be independently proven before releasing the original neutral tool gate");
await release();
}
export function gradePostStop(value: unknown) {
const c = checks(), o = runtimeObservation(value);
c.add("actual-same-source-stop-continuation-evidence", !!o, "A real setup human request, actual native provider cleanup, canonical human instruction and normal continuation must be independently observed."); if (!o) return c.result();
const s = o.state;
c.add("human-stop-cleanup-and-exact-message-chronology", s.stopStatus === 200 && s.messageAcceptedDuringHeldCleanup === true && s.noSuccessorWhileHeld === true && s.gateReleasedAfterCleanup === true && s.chronologyBound === true && s.providerExited === true && s.initialHumanRetained === true,
"The same actual source run has a real Stop intent and ACK; the owned protocol endpoint has retired (the OpenCode proxy, not SDK server health), and only its eligible successor consumes the unchanged canonical message after cleanup.");
c.add("one-eligible-current-owner-with-no-revival-or-duplicate", o.runs.length === 2 && o.runs.filter(row => row.status === "cancelled").length === 1 && o.runs.filter(row => row.status === "succeeded").length === 1
&& o.runs.every(row => row.agentId === o.leadId) && s.singleSuccessor === true && s.sameCurrentAssignment === true && s.oldEvidenceRetained === true && o.humanQueries.length === 2 && o.tasks.length === 0,
"The original stopped execution remains cancelled, one native source continuation owns the new direction under current assignment, and no duplicate task or user message appears.");
c.add("actual-visible-single-page-result-from-continuation", s.actualSimplePage === true && s.artifactRunBound === true && String(s.canonicalReply ?? "").trim().length > 0,
"The subsequent native reply reports the real saved simple single page; registered bytes and visible design independently prove the result.");
return c.result();
}
export const runtimePostStopCases = [runtimeDefinition("recovery.post-stop-message", async ctx => {
const source = await openStandingChat(ctx), lead = await ctx.input.api.get<RuntimeAgent>(`/api/agents/${ctx.input.fixtures.agent.id}`);
const { gate } = await prepareRuntimeWorkerGate(ctx, { worker: lead, taskBound: true });
const secondary = await ctx.input.page.context().newPage(); handles.set(ctx, { page: secondary });
try {
const first = `Build an Acorn Notes website with a visible heading, product paragraph and labelled signup action. Initially use several panels and tabs. Work directly in this conversation; save and register the actual website.html artifact here, without delegating another task. First run the bounded command ${gate.command} once and wait for its output before implementing the website. Do not replace the command or invent its output. Preserve later human changes to this same website.`;
await sendChatMessage(ctx.input.page, first);
const running = await pollUntil({ label: "Genuine initial source Website request executing its bounded native tool", deadlineAt: Date.now() + 150000,
load: async () => ({ runs: await ctx.allRuns(), waiting: await readRuntimeTaskGate(ctx, gate, source.id, lead.id) }),
accept: v => v.waiting === "waiting" && v.runs.length === 1 && v.runs[0]?.agentId === lead.id && v.runs[0].status === "running",
reject: v => v.runs.some(row => row.status !== "running" && row.status !== "queued") ? "The genuine initial source execution ended before Stop setup" : undefined });
const original = await ctx.input.api.get<NativeFaultRun>(`/api/heartbeat-runs/${running.runs[0]!.id}`);
if (original.nativeIssueId !== source.id || original.runtimeMode !== "native") throw new Error("The initial native execution is not bound to the actual canonical source chat");
const people = await ctx.input.api.get<{ user: { id: string } }>("/api/auth/get-session");
const comments = await ctx.input.api.get<IssueComment[]>(`/api/issues/${source.id}/comments?order=asc`), initial = comments.filter(row => !row.authorAgentId);
if (initial.length !== 1 || initial[0]?.body !== first || initial[0].authorUserId !== people.user.id) throw new Error("The initial Website request is not the exact genuine signed-in human setup turn");
await secondary.goto(ctx.input.page.url(), { waitUntil: "domcontentloaded" }); await expect(secondary.getByTestId("task-chat-composer-stop")).toBeEnabled();
return { gate, sourceId: source.id, workerRunId: original.id, initialHuman: initial[0], humanId: people.user.id };
} catch (error) { if (gate.gate) await writeFile(gate.gate, "The pending fixture tool has been released.", "utf8"); await secondary.close(); handles.delete(ctx); throw error; }
}, gradePostStop, {
oracleVersion: "paperclip.chat-catalog.post-stop.v4", canonicalBoundary: "native-stop-cleanup", authentication: "human-session", setupHumanMessageCount: 1, expectedHumanMessageCount: 2, expectedRunCount: 2, attemptTimeoutMs: 720000,
runPolicy: { kind: "terminal-counts", terminalStatusCounts: { cancelled: 1, succeeded: 1 } },
sourceFiles: ["chat-catalog-domains/runtime-post-stop.ts", "chat-catalog-domains/runtime-provider-cleanup.ts", "chat-catalog-domains/runtime-native-fault.ts", "chat-catalog-domains/runtime-busy-worker.ts", "chat-catalog-domains/runtime-rendered-artifact.ts", "chat-catalog-domains/runtime-support.ts", "chat-native-stop-boundary.ts", "../../server/src/services/native-runtime/native-session-executor.ts", "../../server/src/services/explicit-native-continuation.ts", "../../packages/paperclip-runner/src/live/runnerd-codex-transport.ts", "../../packages/paperclip-runner/runner/crates/runner-core/src/codex_provider.rs", "../../packages/paperclip-runner/runner/crates/runner-core/src/process_supervisor.rs", "registered-artifact.ts", "chat-scenarios/recovery.yaml"],
setupPolicy: { readiness: "preserve-scenario", invariantCheckIds: ["actual-source-native-owner"] },
async verifySetup(ctx, state) { const c = checks(), run = await ctx.input.api.get<NativeFaultRun>(`/api/heartbeat-runs/${state.workerRunId}`);
c.add("actual-source-native-owner", run.status === "running" && run.nativeIssueId === ctx.issue()?.id && run.agentId === ctx.input.fixtures.agent.id && run.runtimeMode === "native", "The actual canonical conversation already has its genuine native setup owner, without fabricated status or response history."); return c.result(); },
async beforeCanonicalSend(ctx, state) {
const handle = handles.get(ctx); if (!handle || !ctx.issue()) throw new Error("Actual same-source human Stop tab is unavailable");
await suspendOwnedProviderForCleanup(ctx, { runId: String(state.workerRunId), issueId: ctx.issue()!.id, agentId: ctx.input.fixtures.agent.id });
const endpoint = `/api/heartbeat-runs/${state.workerRunId}/cancel`;
const response = handle.page.waitForResponse(r => r.request().method() === "POST" && new URL(r.url()).pathname === endpoint, { timeout: 30000 });
handle.completed = response.then(r => r.status()); void handle.completed.catch(() => undefined);
await handle.page.getByTestId("task-chat-composer-stop").click();
const run = await pollUntil({ label: "Actual public human Stop durably records pending native cleanup intent", deadlineAt: Date.now() + 1000, intervalMs: 25,
load: () => ctx.input.api.get<NativeFaultRun>(`/api/heartbeat-runs/${state.workerRunId}`),
accept: row => (row.resultJson?.nativeCancellation as StopIntent | undefined)?.dispatchState === "pending" });
const intent = run.resultJson!.nativeCancellation as StopIntent;
if (!intent.intentId || !intent.recordedAt || !intent.intentAuditId || intent.issueId !== ctx.issue()!.id) throw new Error("The actual pending Stop lacks its native source/audit identity");
return { runId: run.id, issueId: ctx.issue()!.id, intentId: intent.intentId, intentRecordedAt: intent.recordedAt };
},
async act(ctx, state) {
// Deterministic cleanup gate: the old provider stays stopped (so its cleanup cannot finish)
// until the follow-up message is durably accepted. Only then is the hold released.
const accepted = await pollUntil({ label: "Follow-up message durably accepted while the stopped run's provider is still held in cleanup", deadlineAt: Date.now() + 60000, intervalMs: 100,
load: async () => ({ comments: await ctx.input.api.get<IssueComment[]>(`/api/issues/${ctx.issue()!.id}/comments?order=asc`), held: await ownedCleanupProviderIsHeld(ctx) }),
accept: v => v.comments.filter(row => !row.authorAgentId).length === 2 && v.comments.filter(row => !row.authorAgentId)[1]?.body === ctx.scenario.query,
reject: v => !v.held ? "The old provider left the fixture's cleanup hold before the follow-up message was accepted" : undefined });
const runsWhileHeld = await ctx.allRuns();
state.noSuccessorWhileHeld = runsWhileHeld.filter(row => row.id !== state.workerRunId).every(row => row.status === "queued" && !(row as { startedAt?: string | null }).startedAt);
state.messageAcceptedDuringHeldCleanup = accepted.held && await ownedCleanupProviderIsHeld(ctx);
if (state.messageAcceptedDuringHeldCleanup !== true) throw new Error("The follow-up message was not accepted inside the held cleanup interval");
state.heldReleasedAt = new Date().toISOString();
await releaseOwnedCleanupProvider(ctx);
const cleaned = await pollUntil({ label: "Actual old source cancellation ACK and owned provider endpoint retirement before neutral gate release", deadlineAt: Date.now() + 90000,
load: async () => ({ old: await ctx.input.api.get<NativeFaultRun>(`/api/heartbeat-runs/${state.workerRunId}`), providerExited: await ownedCleanupProviderHasExited(ctx) }),
accept: v => v.old.status === "cancelled" && (v.old.resultJson?.nativeCancellation as StopIntent | undefined)?.dispatchState === "acknowledged" && v.providerExited });
await releasePostStopGateAfterCleanup(cleaned.old, cleaned.providerExited, () => writeFile((state.gate as { gate: string }).gate, "The pending fixture tool has been released.", "utf8"));
state.gateReleasedAfterCleanup = true;
const settled = await pollUntil({ label: "One normal same-source continuation after actual stopped-provider cleanup", deadlineAt: Date.now() + 300000,
load: async () => { await ctx.refreshIssue(); return { runs: await ctx.allRuns(), source: ctx.issue() }; },
accept: v => v.source?.conversationState === "waiting" && v.runs.length === 2 && v.runs.find(row => row.id === state.workerRunId)?.status === "cancelled" && v.runs.find(row => row.id !== state.workerRunId)?.status === "succeeded",
reject: v => v.runs.length > 2 || v.runs.some(row => row.id !== state.workerRunId && ["failed", "timed_out", "cancelled"].includes(row.status)) ? "The actual continuation duplicated or failed" : undefined });
state.stopStatus = await handles.get(ctx)!.completed;
const old = await ctx.input.api.get<NativeFaultRun>(`/api/heartbeat-runs/${state.workerRunId}`), successor = settled.runs.find(row => row.id !== old.id)!;
const next = await ctx.input.api.get<NativeFaultRun>(`/api/heartbeat-runs/${successor.id}`), comments = await ctx.input.api.get<IssueComment[]>(`/api/issues/${ctx.issue()!.id}/comments?order=asc`);
const human = comments.filter(row => !row.authorAgentId), canonical = human[1];
const native = next.runnerProfileJson?.nativeExecutionInput as { task?: { prompt?: string } } | undefined;
state.chronologyBound = !!canonical && sameSourcePostStopChronology({ boundary: state.nativeStopBoundary as NativeStopBoundaryReceipt, old, next, human: canonical, query: ctx.scenario.query, companyId: ctx.input.fixtures.company.id, agentId: ctx.input.fixtures.agent.id, humanId: String(state.humanId), prompt: native?.task?.prompt ?? "", wakeCommentIds: (next.contextSnapshot?.wakeCommentIds as string[] | undefined) ?? [], heldReleasedAt: String(state.heldReleasedAt) });
state.initialHumanRetained = human[0]?.id === (state.initialHuman as IssueComment).id && human[0]?.body === (state.initialHuman as IssueComment).body;
state.singleSuccessor = settled.runs.filter(row => row.id !== old.id).length === 1;
state.sameCurrentAssignment = ctx.issue()?.assigneeAgentId === next.agentId || (ctx.issue() as { conversationAgentId?: string }).conversationAgentId === next.agentId;
const events = await collectRunEvents<{ seq?: number; eventType?: string; createdAt?: string; payload?: { schema?: string; runId?: string; nativeSessionId?: string } }>((afterSeq, limit) => ctx.input.api.get(`/api/heartbeat-runs/${old.id}/events?afterSeq=${afterSeq}&limit=${limit}`));
state.oldEvidenceRetained = events.length > 0 && old.status === "cancelled";
state.providerExited = await ownedCleanupProviderHasExited(ctx);
if (old.errorCode === "native_session_cleanup_quarantined") state.chronologyBound = state.chronologyBound === true && !!next.nativeSessionId && next.nativeSessionId !== old.nativeSessionId && events.some(event => event.eventType === "native.stopped_conversation_verified" && event.payload?.schema === "paperclip.stopped_native_conversation.v1" && event.payload.runId === old.id && event.payload.nativeSessionId === old.nativeSessionId && Date.parse(event.createdAt ?? "") <= Date.parse(String(next.startedAt)));
state.canonicalReply = comments.filter(row => row.createdByRunId === next.id && row.authorAgentId === next.agentId).map(row => row.body).join("\n"); state.successorRunId = next.id;
await releaseOwnedCleanupProvider(ctx);
},
async observe(ctx, state) {
const o = await observeRuntime(ctx, state);
if (state.successorRunId) { const [artifact] = await readRegisteredArtifacts(ctx.input.api, String(state.sourceId), String(state.successorRunId), ["website.html"]);
state.artifactRunBound = artifact?.contentVerified === true;
const nodes = artifact?.contentVerified ? await renderRuntimeArtifact(ctx, artifact.content) : [];
state.actualSimplePage = !!artifact?.contentVerified && hasSimplerWebsite(nodes);
await ctx.input.evidence("runtime-post-stop-artifact.json", { contentVerified: artifact?.contentVerified === true, bytes: artifact ? Buffer.byteLength(artifact.content) : 0,
checks: simplerWebsiteChecks(nodes), visibleTags: [...new Set(nodes.map(node => node.tag))].sort(), content: artifact?.content ?? null }); }
return o;
},
async verifyResult(ctx) { await ctx.input.page.reload({ waitUntil: "domcontentloaded" }); await expect(ctx.input.page.getByTestId("task-chat-composer-input")).toBeVisible(); },
async cleanup(ctx, state) { try { await releaseOwnedCleanupProvider(ctx); } finally { if (state.gate) await writeFile((state.gate as { gate: string }).gate, "The pending fixture tool has been released.", "utf8"); const handle = handles.get(ctx); if (handle) { await handle.completed?.catch(() => undefined); await handle.page.close(); handles.delete(ctx); } } },
})];