Files
PaperClipAI/tests/runner-e2e/chat-catalog-domains/external-team-install.test.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

233 lines
13 KiB
TypeScript

import type { ChatCatalogContext } from "../chat-catalog-contract.js";
import type { Agent } from "../../../packages/shared/src/types/agent.js";
import { CHAT_OPENCODE_MODEL } from "../chat-runtime-profile.js";
import { updateAgentPermissionsSchema } from "../../../packages/shared/src/validators/agent.js";
import { catalogTeamInstallSchema } from "../../../packages/shared/src/validators/teams-catalog.js";
import { describe, expect, it, vi } from "vitest";
import { teamInstallExternalCases, selectedTeamInstallRuntime, teamInstallOverrides, verifyInstalledTeamRuntime } from "./external-team-install.js";
function valid() {
const team = {
id: "team",
contentHash: "hash",
agentSlugs: ["cto", "coder"],
counts: { agents: 2, projects: 1, routines: 1 },
};
const agent = (id: string) => ({
id,
companyId: "company",
adapterType: "paperclip_runner",
adapterConfig: { provider: "opencode", model: CHAT_OPENCODE_MODEL, timeoutSec: 0, env: { OPENROUTER_API_KEY: { type: "secret_ref", secretId: "account", version: "latest" } } },
urlKey: id,
metadata: {
paperclip: { catalogTeam: { catalogId: "team", originHash: "hash" } },
},
});
const before = { agents: [], projects: [], routines: [], installed: [] };
const after = {
agents: [agent("cto"), agent("coder")],
projects: [{ id: "project", companyId: "company", urlKey: "website" }],
routines: [{ id: "routine", companyId: "company" }],
installed: [{ catalogId: "team", agentCount: 2, outOfDate: false }],
};
return {
scenarioId: "teams.install-team",
origin: "http://fixture.invalid",
companyId: "company",
leadId: "ceo",
phaseReplies: [
{
runId: "initial",
body: "The existing CTO name conflicted, so the team got CTO 2. Team resources were installed, but launch-checklist skill could not be installed because its slug conflicts with the local checklist.",
},
{
runId: "repeat",
body: "The existing team is already installed; I kept its same resources. The launch-checklist skill is still blocked by the local slug conflict.",
},
],
connections: [],
profiles: [],
secrets: [],
endpoints: [],
interactions: [],
tasks: [],
projects: after.projects,
goals: [],
comments: [],
documents: [],
revisions: [],
humanQueries: [],
reply: "Installed the team; it already exists.",
links: [
"/ACORN/agents/cto",
"/ACORN/agents/coder",
"/ACORN/projects/website",
],
runs: [
{ id: "initial", status: "succeeded" },
{ id: "repeat", status: "succeeded" },
],
activity: [
{
id: "install",
action: "company.team_catalog_installed",
actorType: "agent",
agentId: "ceo",
runId: "initial",
details: {
catalogId: "team",
warningCount: 1,
skillPreparationCount: 1,
},
},
],
state: {
blockedSkill: {
id: "required",
key: "catalog-required",
slug: "launch-checklist",
name: "Launch Checklist",
},
renamePlan: { plannedName: "CTO 2" },
actualPreviewCollision: true,
actualSkillCollision409: true,
firstConditional: {
existingAgentPreserved: true,
existingLocalSkillPreserved: true,
requiredCatalogStillMissing: true,
localSkillStillOccupiesSlug: true,
},
finalConditional: {
existingAgentPreserved: true,
existingLocalSkillPreserved: true,
requiredCatalogStillMissing: true,
localSkillStillOccupiesSlug: true,
},
team,
selectedRuntime: { profileId: "runner-opencode", adapterType: "paperclip_runner", provider: "opencode", model: CHAT_OPENCODE_MODEL, accountRef: { type: "secret_ref", secretId: "account", version: "latest" } },
runtimeReadback: { memberIds: ["coder", "cto"], provider: "opencode", model: CHAT_OPENCODE_MODEL, verifiedBeforeImportedWork: true },
before,
first: structuredClone(after),
after,
companyPrefix: "ACORN",
previewAccepted: true,
previewErrorCount: 0,
actorCanCreateAgents: true,
existingAccountReady: true,
originalHumanQuery: "Install that website team in this company.",
repeatHumanQuery: "Install that website team in this company.",
initialRunId: "initial",
},
};
}
describe("real catalog installer oracle", () => {
const definition = teamInstallExternalCases[0]!;
it("accepts attributed installation and actual repeat reconciliation", () => {
expect(definition.grade(valid()).passed).toBe(true);
});
it("rejects plausible duplicate installations, wrong provenance, human prepopulation and substituted repeat transcript", () => {
for (const change of [
(o: ReturnType<typeof valid>) => {
o.state.after.agents.push({ ...o.state.after.agents[0]!, id: "copy" });
},
(o: ReturnType<typeof valid>) => {
o.state.after.agents[0]!.metadata.paperclip.catalogTeam.originHash =
"other-package";
},
(o: ReturnType<typeof valid>) => {
o.activity[0]!.actorType = "user";
},
(o: ReturnType<typeof valid>) => {
o.phaseReplies[1]!.runId = "initial";
},
(o: ReturnType<typeof valid>) => {
o.links[2] = "https://other.invalid/ACORN/projects/project";
},
(o: ReturnType<typeof valid>) => {
o.links[2] = "/OTHER/projects/project";
},
(o: ReturnType<typeof valid>) => {
o.links[2] = "/ACORN/projects";
},
]) {
const o = valid();
change(o);
expect(definition.grade(o).passed).toBe(false);
}
});
it("missing evidence or a successful-sounding reply alone cannot pass", () => {
expect(definition.grade(undefined).passed).toBe(false);
expect(definition.grade({ reply: "Installed everything." }).passed).toBe(
false,
);
});
});
function runtimeFixture() {
const o = valid();
const ctx = { input: { execution: { profile: { id: "runner-opencode", provider: "opencode", model: CHAT_OPENCODE_MODEL, credential: "OPENROUTER_API_KEY" } }, fixtures: { company: { id: "company" }, secretRefs: { OPENROUTER_API_KEY: { type: "secret_ref", secretId: "account", version: "latest" } } } } } as unknown as ChatCatalogContext;
const actor = { ...o.state.after.agents[0]!, id: "ceo", adapterConfig: { ...o.state.after.agents[0]!.adapterConfig, cwd: "/not-a-team-cwd", instructionsFilePath: "/not-a-team-instruction", env: { ...o.state.after.agents[0]!.adapterConfig.env, UNRELATED_VALUE: "must-not-copy" } } } as unknown as Agent;
return { o, ctx, actor };
}
it("the actual public per-slug override schema accepts only the selected runtime/account configuration, not CEO instructions or cwd", () => {
const f = runtimeFixture(), runtime = selectedTeamInstallRuntime(f.ctx, f.actor), overrides = teamInstallOverrides(f.o.state.team as never, runtime);
const payload = catalogTeamInstallSchema.parse({ targetManagerAgentId: "ceo", adapterOverrides: overrides });
expect(Object.keys(payload.adapterOverrides!)).toEqual(["cto", "coder"]);
for (const entry of Object.values(payload.adapterOverrides!)) {
expect(entry).toEqual({ adapterType: "paperclip_runner", adapterConfig: { provider: "opencode", model: CHAT_OPENCODE_MODEL, timeoutSec: 0, lifecycleMode: "per_turn", opencodePermissionMode: "allow", env: { OPENROUTER_API_KEY: { type: "secret_ref", secretId: "account", version: "latest" } } } });
}
expect(JSON.stringify(payload)).not.toContain("must-not-copy"); expect(JSON.stringify(payload)).not.toContain("not-a-team");
});
it.each(["provider", "model", "account", "plaintext", "company"])("rejects a wrong selected %s before exposing installer context", invalid => {
const f = runtimeFixture();
if (invalid === "provider") f.actor.adapterConfig.provider = "codex";
if (invalid === "model") f.actor.adapterConfig.model = "openrouter/other/model";
if (invalid === "account") (f.actor.adapterConfig.env as Record<string, unknown>).OPENROUTER_API_KEY = { type: "secret_ref", secretId: "foreign", version: "latest" };
if (invalid === "plaintext") (f.actor.adapterConfig.env as Record<string, unknown>).OPENROUTER_API_KEY = "must-not-copy";
if (invalid === "company") f.actor.companyId = "foreign";
expect(() => selectedTeamInstallRuntime(f.ctx, f.actor)).toThrow("actual configured");
});
it("independently reads every actual member before another human request or imported native work", async () => {
const f = runtimeFixture(), seen: string[] = [];
Object.assign(f.ctx.input, { api: { get: async (route: string) => { seen.push(route); return f.o.state.after.agents.find(agent => route.endsWith(agent.id)); } } });
Object.assign(f.ctx, { allRuns: async () => [{ id: "source", agentId: "ceo", status: "succeeded", startedAt: "2026-10-28T01:00:00Z" }] });
const proof = await verifyInstalledTeamRuntime(f.ctx, f.o.state.team as never, f.o.state.first as never, selectedTeamInstallRuntime(f.ctx, f.actor));
expect(seen).toEqual(["/api/agents/cto", "/api/agents/coder"]); expect(proof).toMatchObject({ memberIds: ["coder", "cto"], verifiedBeforeImportedWork: true, model: CHAT_OPENCODE_MODEL });
});
it.each(["model", "account", "wrong returned identity", "already started"])("fails actual member runtime readback for %s", async invalid => {
const f = runtimeFixture();
const saved = structuredClone(f.o.state.after.agents);
if (invalid === "model") saved[0]!.adapterConfig.model = "openrouter/other/model";
if (invalid === "account") saved[0]!.adapterConfig.env.OPENROUTER_API_KEY.secretId = "foreign";
if (invalid === "wrong returned identity") saved[0]!.id = "other";
Object.assign(f.ctx.input, { api: { get: async (route: string) => saved[route.endsWith("cto") ? 0 : 1] } });
Object.assign(f.ctx, { allRuns: async () => invalid === "already started" ? [{ id: "work", agentId: "cto", status: "running", startedAt: "2026-10-28T01:00:00Z" }] : [] });
await expect(verifyInstalledTeamRuntime(f.ctx, f.o.state.team as never, f.o.state.first as never, selectedTeamInstallRuntime(f.ctx, f.actor))).rejects.toThrow();
});
it("real resource counts and actor attribution cannot conceal wrong member providers/models/accounts or missing pre-work readback", () => {
for (const change of [
(o: ReturnType<typeof valid>) => { o.state.after.agents[0]!.adapterConfig.provider = "codex"; },
(o: ReturnType<typeof valid>) => { o.state.first.agents[0]!.adapterConfig.model = "openrouter/other/model"; },
(o: ReturnType<typeof valid>) => { o.state.after.agents[0]!.adapterConfig.env.OPENROUTER_API_KEY.secretId = "foreign"; },
(o: ReturnType<typeof valid>) => { o.state.runtimeReadback.verifiedBeforeImportedWork = false; },
(o: ReturnType<typeof valid>) => { o.state.runtimeReadback.memberIds = ["other", "cto"]; },
(o: ReturnType<typeof valid>) => { o.state.selectedRuntime.model = "openrouter/other/model"; o.state.after.agents.forEach(agent => { agent.adapterConfig.model = "openrouter/other/model"; }); },
]) { const o = valid(); change(o); expect(teamInstallExternalCases[0]!.grade(o).passed).toBe(false); }
});
it.each([true,false])("sends the complete permission schema without changing effective assignment=%s",async canAssignTasks=>{
const actor={id:"ceo",companyId:"company",role:"ceo",access:{canAssignTasks}};
const patch=vi.fn(async(_route:string,body:unknown)=>{expect(updateAgentPermissionsSchema.parse(body)).toEqual({canCreateAgents:true,canAssignTasks});throw new Error("stop-after-validated-permission-request");});
const ctx={input:{fixtures:{company:{id:"company"},agent:{id:"ceo"}},api:{get:async(route:string)=>route==="/api/teams/catalog"?[{slug:"product-engineering",compatibility:"compatible"}]:actor,patch}}} as unknown as ChatCatalogContext;
await expect(teamInstallExternalCases[0]!.setup(ctx)).rejects.toThrow("stop-after-validated-permission-request");
expect(patch).toHaveBeenCalledOnce();
});
it.each(["foreign company","wrong agent","not CEO","missing effective assignment"])("refuses %s identity before permission mutation",async invalid=>{
const actor={id:"ceo",companyId:"company",role:"ceo",access:{canAssignTasks:true}};
if(invalid==="foreign company")actor.companyId="foreign";
if(invalid==="wrong agent")actor.id="other";
if(invalid==="not CEO")actor.role="engineer";
if(invalid==="missing effective assignment")delete(actor as {access?:unknown}).access;
const patch=vi.fn(); const ctx={input:{fixtures:{company:{id:"company"},agent:{id:"ceo"}},api:{get:async(route:string)=>route==="/api/teams/catalog"?[{slug:"product-engineering",compatibility:"compatible"}]:actor,patch}}} as unknown as ChatCatalogContext;
await expect(teamInstallExternalCases[0]!.setup(ctx)).rejects.toThrow("observed installer");expect(patch).not.toHaveBeenCalled();
});