Files
PaperClipAI/tests/runner-e2e/chat-catalog-domains/runtime-private-key-worker.ts
T
f4337bb5fb test(evals): land the chat catalog eval harness on master (#19614)
## Thinking Path

> - Paperclip is the open source app people use to manage AI agents for
work
> - The chat-feature evals measure whether agents complete real chat
requests through the product
> - The chat catalog eval harness and its repairs live on the
`chat-eval/master-candidate` branch, not on master
> - Today's work made 139 of 242 chat evals pass in verified single-case
live runs. Each product fix landed on master as its own pull request
> - The eval-side changes (setups, graders, fault seams, evidence
capture) still exist only on the candidate
> - This pull request lands the harness on master. Master stays the
source of truth for product behavior
> - The benefit is one tree: evals run against master commits, and
product fixes reach the evals as soon as they merge

## Linked Issues or Issue Description

Refs #19441, Refs #19599, Refs #20270, Refs #20334, Refs #20342, Refs
#20344, Refs #20354, Refs #20360, Refs #20373, Refs #20382, Refs #20409

**What existing behavior does this improve?**
The chat catalog eval harness and the product it evaluates live on a
long-running branch outside master. Eval fixes do not reach master, so
master-based eval runs use old setups and graders.

**Subsystem affected**
`tests/runner-e2e` chat catalog harness, plus a small set of product
changes that exist only on the candidate.

**Proposed change**
Land the harness on master. Every product fix from this eval work
already merged to master as its own pull request. Conflicts keep
master's product behavior.

## What Changed

- Most of the diff (808 of 851 files) is under `tests/runner-e2e/`:
chat-catalog domains, scenarios, descriptors, the TestOps manifest,
private evidence tooling, and their tests.
- Graders now check saved state, setup literals and rendered or clicked
links. Many keyword and regex checks on free-form replies are removed.
Each changed grader has a bumped `oracleVersion` and calibration tests
(a valid reply, a reworded valid reply, and broken variants that must
fail).
- The shared link helper `hasLink` no longer treats a list link (for
example `/COM/issues?view=mine`) as a link to every task without a URL
key.
- Setup fixes build real state through supported product routes: routine
history, agent grants through the board grant route, budget holds with a
real cap, and a storage-layer archive fault for
`inbox-search.archive-partial-failure`.
- Browser boundary journeys (sign-in, chat-disabled, onboarding) now
write the standard pass evidence files.
- Retargeted cases, approved by the maintainer:
`agent-chat.chat-disabled` (Conference Room gate),
`agents.restore-instructions` (no saved instruction history),
`access.manage-membership` (member permission edits stay board-only).
- The private evidence trusted files (`manual-private-handoff.ts`,
`manual-private-retention.ts`, `manual-private-report-assets.ts`) are
byte-identical to master.
- Product files outside `tests/` (43 files, +1,328 −49) carry
candidate-only changes that master does not yet have. Most are tests.
The main code changes are in `request-human-input.ts`,
`heartbeat-run-summary.ts`, `native-completion-feedback.ts`,
`native-run-finalizer.ts`, `openapi.ts`, `secrets.ts` and
`IssueChatLongThreadPerf.tsx`.

## Verification

- `pnpm vitest run chat-catalog-domains chat-catalog-contract.test.ts
chat-catalog-consistency.test.ts unit-launcher.test.mjs
--testTimeout=180000` in `tests/runner-e2e`: 269 files pass, 3,550 tests
pass.
- `node --test chat-full-catalog-testops-manifest.test.mjs`: 10 pass.
- `node tests/runner-e2e/source-sdk-public-inputs.mjs <HEAD> --check`:
no change.
- `tsc` for `server` and `tests/runner-e2e`: clean.
- Live single-case verification: 139 of 242 cases passed on paid runs,
and a person read the saved evidence for each. These are single-case
runs, not one full-cohort run.

## Risks

- Large diff. Most of it is test-only, but the 43 non-test files change
product code paths. Review those files first.
- Some graders are now less strict on reply wording. A person must read
replies for meaning on those cases.
- 42 cases need disposable outside accounts (Notion, Slack, GitHub,
Daytona and others) and stay blocked. `access.invite-teammate` stays
failing until local agent isolation lands, because the runner's durable
state keeps a live invite token.
- The paid workflow compares the trusted private-evidence files with
master. They are unchanged, so this risk is low.

## Model Used

Anthropic Claude Opus 5.5 (`claude-opus-5-5`), with tool use and code
execution, in Claude Code. Subagents used the same model.

## 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>
Co-authored-by: Paperclip <noreply@paperclip.ing>
2026-11-10 08:02:11 -06:00

92 lines
8.9 KiB
TypeScript

import { readFile } from "node:fs/promises";
import { chromium, test, type Cookie } from "@playwright/test";
import { containsPrivateCredential, type PrivateKeyReceipt } from "./runtime-private-key.js";
import { collectRunEvents } from "../run-observations.js";
interface Request { mode: "agent-key" | "canary"; baseURL: string; profileURL?: string; companyId?: string; agentId?: string; sourceIssueId?: string; cookies: Cookie[]; canary?: string; executablePath?: string; channel?: string }
async function main(): Promise<PrivateKeyReceipt> {
let attached = false; try { test.info(); attached = true; } catch { /* expected in standalone process */ }
if (attached) throw new Error("Private key worker cannot execute in Playwright Test");
const input = JSON.parse(await readFile(process.argv[2]!, "utf8")) as Request, base = new URL(input.baseURL);
if (base.hostname !== "127.0.0.1" || base.protocol !== "http:" || !["agent-key", "canary"].includes(input.mode)) throw new Error("Private key target is not the isolated loopback instance");
const receipt: PrivateKeyReceipt = { privateBrowserProcessIsolated: true, privateBrowserTestRunnerAttached: false, privateBrowserApiStepRecorder: false };
const browser = await chromium.launch({ headless: true, ...(input.executablePath ? { executablePath: input.executablePath } : {}), ...(input.channel ? { channel: input.channel } : {}) });
let keyId: string | undefined;
const cookie = input.cookies.map(row => `${row.name}=${row.value}`).join("; ");
async function response(suffix: string, options: RequestInit = {}) {
const target = new URL(suffix, base); if (target.origin !== base.origin || !target.pathname.startsWith("/api/")) throw new Error("Private scan escaped its instance");
return fetch(target, { redirect: "error", signal: AbortSignal.timeout(15_000), ...options, headers: { Cookie: cookie, Origin: base.origin, ...options.headers } });
}
async function json<T>(suffix: string): Promise<T> { const result = await response(suffix); if (!result.ok) { await result.body?.cancel(); throw new Error("Private public-resource scan failed"); } return result.json() as Promise<T>; }
try {
const context = await browser.newContext({ baseURL: base.origin }); await context.addCookies(input.cookies);
const page = await context.newPage(); page.setDefaultTimeout(30_000);
if (input.mode === "canary") {
if (!input.canary || input.canary.length > 8192) throw new Error("Missing finite private canary");
await page.setContent('<code></code><input type="password" aria-label="Synthetic private canary">');
await page.getByLabel("Synthetic private canary").fill(input.canary);
await page.locator("code").evaluate((node, value) => { node.textContent = value; }, input.canary);
receipt.canaryContained = await page.locator("code").textContent() === input.canary;
return receipt;
}
if (![input.agentId, input.companyId, input.sourceIssueId].every(id => typeof id === "string" && /^[a-f0-9-]{36}$/i.test(id)) || !input.profileURL || new URL(input.profileURL).origin !== base.origin) throw new Error("Invalid exact private key resource binding");
// The profile replaces an id route with the agent's urlKey and remounts the key form, which drops a name typed before then.
const { urlKey } = await json<{ urlKey: string }>(`/api/agents/${input.agentId}`);
await page.goto(input.profileURL, { waitUntil: "domcontentloaded" });
await page.waitForURL(url => url.pathname.split("/").map(decodeURIComponent).join("/").endsWith(`/agents/${urlKey}/api-keys`));
const keyName = page.getByPlaceholder("Key name (e.g. production)");
for (let attempt = 1; ; attempt++) {
await keyName.fill("local integration"); await page.waitForTimeout(750);
if (await keyName.inputValue() === "local integration") break;
if (attempt >= 5) throw new Error("Key name input did not keep its value");
}
await page.getByRole("button", { name: "Create", exact: true }).click();
await page.getByText("API key created — copy it now, it will not be shown again.", { exact: true }).waitFor({ state: "visible" });
const token = (await page.locator("code").filter({ hasText: /^pcp_/ }).first().textContent())?.trim();
if (!token || token.length < 20) throw new Error("Missing private one-time credential");
const identity = await fetch(new URL("/api/agents/me", base), { redirect: "error", signal: AbortSignal.timeout(15_000), headers: { Authorization: `Bearer ${token}` } });
if (!identity.ok) { await identity.body?.cancel(); throw new Error("Human credential authentication failed"); }
const actor = await identity.json() as { id: string; companyId: string };
const keys = await json<Array<{ id: string; name: string; revokedAt: string | null }>>(`/api/agents/${input.agentId}/keys`), created = keys.filter(row => row.name === "local integration" && !row.revokedAt);
if (created.length !== 1) throw new Error("Human key operation did not create exactly one key");
keyId = created[0]!.id; receipt.createdKey = { id: keyId, name: "local integration", revokedAt: null };
receipt.createdKeyAgentId = actor.id; receipt.createdKeyCompanyId = actor.companyId; receipt.humanCreated = true; receipt.tokenShownOnce = true;
await page.getByRole("button", { name: "Dismiss", exact: true }).click(); await page.reload({ waitUntil: "domcontentloaded" });
receipt.tokenAbsentAfterReload = !(await page.locator("body").textContent())?.includes(token);
await page.getByText("local integration", { exact: true }).locator("..").locator("..").getByRole("button", { name: "Revoke", exact: true }).click();
await page.getByText("Revoked Keys", { exact: true }).waitFor({ state: "visible" });
const denied = await fetch(new URL("/api/agents/me", base), { redirect: "error", signal: AbortSignal.timeout(15_000), headers: { Authorization: `Bearer ${token}` } });
receipt.revokedCredentialStatus = denied.status; await denied.body?.cancel();
try {
const root = `/api/issues/${input.sourceIssueId}`, parts: unknown[] = [await json(`${root}/comments?order=asc`), await json(`/api/agents/${input.agentId}/keys`), await json(`/api/companies/${input.companyId}/activity?limit=1000`)];
const documents = await json<Array<{ key: string }>>(`${root}/documents`);
for (const doc of documents) { const url = `${root}/documents/${encodeURIComponent(doc.key)}`; parts.push(await json(url), await json(`${url}/revisions`)); }
parts.push(await json(`${root}/work-products`));
const attachments = await json<Array<{ id: string; byteSize: number }>>(`${root}/attachments`); parts.push(attachments);
for (const attachment of attachments) {
if (!Number.isSafeInteger(attachment.byteSize) || attachment.byteSize < 0 || attachment.byteSize > 1_048_576) throw new Error("Private attachment scan exceeded bound");
const content = await response(`/api/attachments/${attachment.id}/content`); if (!content.ok) { await content.body?.cancel(); throw new Error("Private attachment read failed"); }
parts.push(Buffer.from(await content.arrayBuffer()).toString("utf8"));
}
const runs = await json<Array<{ id: string }>>(`/api/companies/${input.companyId}/heartbeat-runs?limit=100`); parts.push(runs);
if (runs.length >= 100) throw new Error("Private run scan requires complete bounded run inventory");
let logBytes = 0;
for (const run of runs) {
parts.push(await json(`/api/heartbeat-runs/${run.id}`)); let offset = 0;
for (;;) {
const log = await json<{ content: string; nextOffset?: number }>(`/api/heartbeat-runs/${run.id}/log?offset=${offset}&limitBytes=1048576`);
if (typeof log.content !== "string") throw new Error("Missing complete private run log");
parts.push(log.content); logBytes += Buffer.byteLength(log.content); if (logBytes > 8_388_608) throw new Error("Private log scan exceeded bound");
if (log.nextOffset === undefined) break; if (log.nextOffset <= offset) throw new Error("Private log pagination failed"); offset = log.nextOffset;
}
parts.push(await collectRunEvents((afterSeq, limit) => json(`/api/heartbeat-runs/${run.id}/events?afterSeq=${afterSeq}&limit=${limit}`)));
}
receipt.tokenLeaked = containsPrivateCredential(parts, token); receipt.privateScanComplete = true; receipt.artifactScanComplete = true; receipt.runEvidenceScanComplete = true;
} catch { receipt.privateScanComplete = false; receipt.artifactScanComplete = false; receipt.runEvidenceScanComplete = false; }
return receipt;
} finally {
if (keyId && input.agentId) { try { const result = await response(`/api/agents/${input.agentId}/keys/${keyId}`, { method: "DELETE" }); await result.body?.cancel(); } catch { /* company cleanup retains final ownership */ } }
await browser.close();
}
}
main().then(value => process.stdout.write(JSON.stringify(value))).catch(() => { process.stderr.write("Private key operation failed; diagnostics withheld"); process.exitCode = 1; });