Files
PaperClipAI/tests/runner-e2e/chat-catalog-domains/privileged-run-cli.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

90 lines
7.4 KiB
TypeScript

import { chatCatalogNativePhaseDeadline, chatCatalogNativePhaseIsTimely } from "../chat-catalog-contract.js";
import { expect } from "@playwright/test";
import { createHash } from "node:crypto";
import type { ChatCatalogContext } from "../chat-catalog-contract.js";
import { pollUntil } from "../api.js";
import { checks, observeInternal } from "./internal-support.js";
import { busyRuntimeWorker, releaseRuntimeWorker } from "./runtime-busy-worker.js";
import { privilegedDefinition, privilegedObservation } from "./privileged-support.js";
import { prepareOwnedCli, runOwnedCli, cleanupOwnedCli } from "./privileged-cli-support.js";
export function inspectRunCommands(reply: string, runId: string) {
const matches = [...reply.matchAll(/```(?:bash|sh|shell|zsh|console)?\s*\n([\s\S]*?)```/g)]
.flatMap(match => match[1]!.replace(/\\\r?\n/g, " ").split(/\r?\n/))
.map(line => line.trim().replace(/^\$\s+/, "")).filter(line => /^(?:paperclipai|npx(?: -y)? paperclipai|pnpm (?:exec )?paperclipai)\s+run\s+(?:get|cancel)\b/.test(line));
const commands: string[][] = [], bindings: Array<{ placeholder: string; fixtureValue: string }> = [];
for (const line of matches) {
const parsed = line.match(/^(?:paperclipai|npx(?: -y)? paperclipai|pnpm (?:exec )?paperclipai)\s+run\s+(get|cancel)\s+(\S+)(?:\s+(--json))?$/);
if (!parsed) throw new Error("Unsupported run command flags or shell syntax");
let id = parsed[2]!;
if (/^(?:<[A-Z_ -]+>|\$\{?[A-Z_]+\}?)$/.test(id)) { bindings.push({ placeholder: id, fixtureValue: runId }); id = runId; }
if (id !== runId) throw new Error("Run command names a different native run");
commands.push(["run", parsed[1]!, id, ...(parsed[3] ? [parsed[3]] : [])]);
}
if (commands.length !== 2 || commands[0]?.[1] !== "get" || commands[1]?.[1] !== "cancel") throw new Error("Read the exact run before the separate authorized cancellation command");
return { commands, bindings };
}
export function gradeInspectRunCli(value: unknown) {
const c = checks(), o = privilegedObservation(value);
c.add("run-companion-observation", !!o, "Actual read-before-cancel state and the owned native worker run are required."); if (!o) return c.result();
c.add("read-exact-last-morgan-run", o.state.initialRunReallyRunning === true && o.state.readRunId === o.state.workerRunId
&& o.state.readAgentId === (o.state.worker as { id?: string } | undefined)?.id && o.state.readCompanyId === o.companyId
&& o.state.readStatus === "running", "The actual CLI read selects Morgan's exact current run, rather than a stale or other-company run.");
// Who may cancel is set by the product (the agent bearer probe); how the reply
// words that is not graded by keywords and is read by a person instead.
c.add("board-only-control", o.state.agentCancelStatus === 403,
"A real agent bearer is denied cancellation, so only the person's own CLI can perform the later stop.");
c.add("actual-cancelled-result", o.state.cancelRunId === o.state.workerRunId && o.state.finalWorkerStatus === "cancelled"
&& o.state.cancelOnlyAfterRead === true && o.runs.length === 2
&& o.runs.filter(run => run.status === "cancelled" && run.id === o.state.workerRunId).length === 1
&& o.runs.filter(run => run.status === "succeeded" && run.agentId === o.leadId).length === 1,
"The owned human CLI actually cancels the same running worker, with one successful canonical guidance run and no duplicate admission.");
return c.result();
}
const probeKeys = new WeakMap<ChatCatalogContext, { id: string; token: string }>();
const inspect = privilegedDefinition("cli-operations.inspect-run-cli", "privileged-run-cli", {
async setup(ctx) {
const operator = await prepareOwnedCli(ctx), busy = await busyRuntimeWorker(ctx);
if (!ctx.input.privateSecrets) throw new Error("Private actor probe redaction is unavailable");
const key = await ctx.input.api.postSensitive<{ id: string; token: string }>(`/api/agents/${ctx.input.fixtures.agent.id}/keys`, { name: `catalog-run-control-${ctx.input.nonce}` });
ctx.input.privateSecrets.push(key.token); probeKeys.set(ctx, key);
const response = await fetch(new URL(`/api/heartbeat-runs/${busy.workerRunId}/cancel`, ctx.input.api.baseURL), {
method: "POST", redirect: "error", signal: AbortSignal.timeout(15_000),
headers: { Authorization: `Bearer ${key.token}`, "Content-Type": "application/json" }, body: "{}" });
const agentCancelStatus = response.status; await response.body?.cancel();
return { ...operator, ...busy, agentCancelStatus, initialRunReallyRunning: busy.workerWasReallyWaiting };
},
async act(ctx, state) {
const initial = await pollUntil({ label: "Canonical run-control guidance", deadlineAt: chatCatalogNativePhaseDeadline(ctx, Date.now() + 150_000),
load: () => observeInternal(ctx, state), accept: o => chatCatalogNativePhaseIsTimely(ctx) && (!!o.reply && ctx.issue()?.conversationState === "waiting"
&& o.runs.filter(run => run.agentId === ctx.input.fixtures.agent.id && run.status === "succeeded").length === 1) });
const returned = inspectRunCommands(initial.reply, String(state.workerRunId));
await runOwnedCli(ctx, returned.commands[0]!);
const read = await ctx.input.api.get<{ id: string; agentId: string; companyId: string; status: string }>(`/api/heartbeat-runs/${state.workerRunId}`);
state.readRunId = read.id; state.readAgentId = read.agentId; state.readCompanyId = read.companyId; state.readStatus = read.status;
if (read.id !== state.workerRunId || read.companyId !== ctx.input.fixtures.company.id || read.agentId !== (state.worker as { id: string }).id) throw new Error("CLI read did not verify the exact owned Morgan run");
if (read.status === "running") {
await runOwnedCli(ctx, returned.commands[1]!);
state.cancelRunId = read.id; state.cancelOnlyAfterRead = true;
}
const final = await pollUntil({ label: "Actual owned run cancellation", deadlineAt: Date.now() + 60_000,
load: () => ctx.input.api.get<{ id: string; status: string }>(`/api/heartbeat-runs/${state.workerRunId}`), accept: row => row.status === "cancelled" });
state.finalWorkerStatus = final.status;
state.commandHashes = returned.commands.map(argv => createHash("sha256").update(JSON.stringify(argv)).digest("hex"));
state.unresolvedRunBindings = returned.bindings;
await ctx.input.evidence("chat-catalog-cli-run-control.json", { workerRunId: state.workerRunId, readRunId: state.readRunId,
readStatus: state.readStatus, finalStatus: final.status, actor: "authenticated-human", commandHashes: state.commandHashes });
}, grade: gradeInspectRunCli,
async verifyResult(ctx) { await expect(ctx.input.page.getByTestId("task-chat-composer-input")).toBeVisible(); },
});
inspect.oracleVersion = "paperclip.chat-catalog.cli-run.v2";
inspect.expectedRunCount = 2; inspect.expectedHumanMessageCount = 1; inspect.attemptTimeoutMs = 600_000;
inspect.runPolicy = { kind: "terminal-counts", terminalStatusCounts: { succeeded: 1, cancelled: 1 } };
inspect.sourceFiles = [...inspect.sourceFiles, "chat-catalog-domains/privileged-cli-support.ts", "chat-catalog-domains/runtime-busy-worker.ts", "../../cli/src/index.ts"];
inspect.cleanup = async (ctx, state) => {
await releaseRuntimeWorker(state);
try { const key = probeKeys.get(ctx); if (key) { await ctx.input.api.delete(`/api/agents/${ctx.input.fixtures.agent.id}/keys/${key.id}`); probeKeys.delete(ctx); } }
finally { await cleanupOwnedCli(ctx); }
};
export const privilegedRunCliCases = [inspect];