Files
PaperClipAI/tests/runner-e2e/chat-private-inputs.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

135 lines
13 KiB
TypeScript

import { CHAT_ORACLE_KEY_ENV, privateOracleIdentity, oracleSemanticFields, verifyRetainedPrivateOracle, type PrivateOracleIdentity } from "./chat-private-oracle.js";
import { createHash } from "node:crypto";
import { lstatSync, readFileSync, realpathSync, writeFileSync } from "node:fs";
import path from "node:path";
import { CREDENTIAL_NAMES, type CredentialName, type MatrixExecution } from "./types.js";
import type { ChatCatalogDescriptor } from "./chat-catalog-registry.js";
import { runtimePrivateProviders, runtimePrivateReferenceNames, validateRuntimePrivateRequirement, type RuntimePrivateProvider } from "./chat-runtime-private-requirements.js";
export const CHAT_PRIVATE_INPUTS_ENV = "PAPERCLIP_CHAT_PRIVATE_INPUTS";
const externalConfigEnv = "PAPERCLIP_CHAT_CATALOG_EXTERNAL_CONFIG";
export interface ChatExternalInputRequirement { provider: "notion" | "slack" | "agentmail" | "vault" | "github" | RuntimePrivateProvider; fields: readonly string[] }
type PrivateInputDescriptor = Pick<ChatCatalogDescriptor, "requiredCredentials" | "externalInput"> & Partial<Pick<ChatCatalogDescriptor, "scenarioId">>;
export interface ChatPrivateInputs {
readonly setupBlockers?: readonly string[];
readonly credentials: Readonly<Partial<Record<CredentialName, string>>>;
read(name: string): string | undefined;
configuration(name: "external-provider"): unknown;
}
interface PrivateEnvelope { version: 1; executionId: string; credentials: Partial<Record<CredentialName, string>>; references: Record<string, string>; external?: unknown; setupBlockers?: string[]; oracleExpectation?: PrivateOracleIdentity; declaredCredentialNames?: string[]; declaredReferenceNames?: string[]; declaredExternalRequirements?: Array<{ provider: string; fields: string[] }> }
function readPrivateJson(filename: string, ownedRoot?: string): unknown {
const info = lstatSync(filename);
if (!info.isFile() || info.isSymbolicLink() || info.mode & 0o077 || info.size > 262144
|| (ownedRoot && !realpathSync(filename).startsWith(realpathSync(ownedRoot) + path.sep)))
throw new Error("Selected-case inputs require a bounded private0600 owned file; contents withheld");
try { return JSON.parse(readFileSync(filename, "utf8")); } catch { throw new Error("Selected-case private inputs are invalid; contents withheld"); }
}
export function chatPrivateInputRequirements(descriptor?: PrivateInputDescriptor) {
const credentialNames = [...new Set(descriptor?.requiredCredentials ?? [])].sort();
if (credentialNames.some(name => !CREDENTIAL_NAMES.includes(name))) throw new Error("Unknown selected-case credential name");
const external = descriptor?.externalInput;
if (external && (!["notion", "slack", "agentmail", "vault", "github", ...runtimePrivateProviders].includes(external.provider)
|| !external.fields.length || external.fields.some(field => !/^[a-zA-Z][a-zA-Z0-9]*$/.test(field)))) throw new Error("Invalid selected external configuration requirements");
if (external) validateRuntimePrivateRequirement(descriptor?.scenarioId, external.provider, external.fields, credentialNames);
const requirements = { credentialNames, external: external ? { provider: external.provider, fields: [...external.fields].sort() } : null };
return { ...requirements, digest: createHash("sha256").update(JSON.stringify(requirements)).digest("hex") };
}
export function stageChatPrivateInputs(input: { execution: MatrixExecution; descriptor?: PrivateInputDescriptor; privateDir: string; env: NodeJS.ProcessEnv; secrets: string[] }) {
if (input.descriptor?.scenarioId && input.execution.task?.id !== input.descriptor.scenarioId) throw new Error("Selected private input declaration belongs to another scenario");
const requirements = chatPrivateInputRequirements(input.descriptor);
const envelope: PrivateEnvelope = { version: 1, executionId: input.execution.id, credentials: {}, references: {}, setupBlockers: [], declaredCredentialNames: requirements.credentialNames, declaredReferenceNames: [], declaredExternalRequirements: requirements.external ? [{ provider: requirements.external.provider, fields: requirements.external.fields }] : [] };
for (const name of requirements.credentialNames) {
const value = input.env[name]?.trim(); if (!value) { envelope.setupBlockers!.push(`missing-credential:${name}`); continue; }
envelope.credentials[name] = value; input.secrets.push(value);
}
if (requirements.external) {
try {
const ref = input.env[externalConfigEnv]; if (!ref) throw new Error("Missing selected-case private external configuration");
const config = readPrivateJson(ref) as { schemaVersion?: number; providers?: Record<string, { disposable?: boolean; resourceIds?: string[]; credentialEnv?: string[]; prerequisites?: Record<string, string> }> };
const binding = config.schemaVersion === 1 ? config.providers?.[requirements.external.provider] : undefined;
if (binding?.prerequisites) for (const field of requirements.external.fields) {
if (typeof binding.prerequisites[field] !== "string" || !binding.prerequisites[field]?.trim()) envelope.setupBlockers!.push(`missing-external-field:${requirements.external.provider}:${field}`);
}
if (!binding || binding.disposable !== true || !Array.isArray(binding.resourceIds) || !binding.resourceIds.length
|| !Array.isArray(binding.credentialEnv) || (!binding.credentialEnv.length && !runtimePrivateProviders.includes(requirements.external.provider)) || !binding.prerequisites
|| requirements.external.fields.some(field => typeof binding.prerequisites?.[field] !== "string" || !binding.prerequisites[field]?.trim())) throw new Error("Selected disposable external configuration lacks declared prerequisites; contents withheld");
const selected = { disposable: true, resourceIds: binding.resourceIds, credentialEnv: binding.credentialEnv,
prerequisites: Object.fromEntries(requirements.external.fields.map(field => [field, binding.prerequisites![field]])) };
const names = [...binding.credentialEnv, ...requirements.external.fields.filter(field => /(?:Env|EnvName)$/.test(field)).map(field => binding.prerequisites![field]!)];
if (runtimePrivateProviders.includes(requirements.external.provider)) {
const expected = runtimePrivateReferenceNames(requirements.external.provider);
if (expected.some(name => !names.includes(name)) || names.some(name => !expected.includes(name))) throw new Error("Runtime private reference names do not match the selected provider's finite contract");
}
envelope.declaredReferenceNames = [...new Set(names)].sort();
for (const name of new Set(names)) {
if (!/^PAPERCLIP_CHAT_EVAL_[A-Z0-9_]+$/.test(name)) throw new Error("External credential reference is not a declared private environment name");
const value = input.env[name]?.trim(); if (!value) { envelope.setupBlockers!.push(`missing-external-reference:${name}`); continue; }
envelope.references[name] = value; input.secrets.push(value);
}
try { envelope.oracleExpectation = privateOracleIdentity(requirements.external.provider, requirements.external.fields, selected.prerequisites, input.env); }
catch { envelope.setupBlockers!.push(`unavailable-oracle-commitment:${requirements.external.provider}`); }
envelope.external = { schemaVersion: 1, providers: { [requirements.external.provider]: selected } };
} catch { envelope.setupBlockers!.push(`unavailable-external-configuration:${requirements.external.provider}`); }
}
const filename = path.join(input.privateDir, "chat-selected-inputs.json");
writeFileSync(filename, JSON.stringify(envelope), { mode: 0o600, flag: "wx" });
return { filename, requirements };
}
export function readChatPrivateInputs(executionId: string, secrets: string[], env: NodeJS.ProcessEnv = process.env): ChatPrivateInputs {
const reference = env[CHAT_PRIVATE_INPUTS_ENV], root = env.PAPERCLIP_RUNNER_E2E_TEMP_ROOT;
if (!reference) return { credentials: Object.freeze({}), read: () => undefined, configuration: () => undefined };
if (!root) throw new Error("Selected private input ownership root is missing");
const envelope = readPrivateJson(reference, root) as PrivateEnvelope;
if (envelope.version !== 1 || envelope.executionId !== executionId || !envelope.credentials || !envelope.references)
throw new Error("Private inputs belong to another execution; contents withheld");
for (const value of [...Object.values(envelope.credentials), ...Object.values(envelope.references)]) {
if (typeof value !== "string" || !value.trim()) throw new Error("Invalid private credential value; contents withheld");
secrets.push(value);
}
if (envelope.setupBlockers?.some(value => !/^(?:missing-credential:[A-Z0-9_]+|missing-external-reference:PAPERCLIP_CHAT_EVAL_[A-Z0-9_]+|missing-external-field:[a-zA-Z]+:[a-zA-Z][a-zA-Z0-9]*|(?:unavailable-external-configuration|unavailable-oracle-commitment):[a-zA-Z]+)$/.test(value))) throw new Error("Invalid private setup blocker receipt");
return { setupBlockers: Object.freeze(envelope.setupBlockers ?? []), credentials: Object.freeze(envelope.credentials), read: name => envelope.references[name] ?? envelope.credentials[name as CredentialName],
configuration: name => name === "external-provider" ? envelope.external : undefined };
}
export function scrubChatPrivateEnvironment(env: NodeJS.ProcessEnv, executions: readonly MatrixExecution[], removeReference = false) {
delete env[CHAT_ORACLE_KEY_ENV];
for (const key of Object.keys(env)) if (key.startsWith("PAPERCLIP_CHAT_EVAL_") || key === externalConfigEnv || removeReference && [CHAT_PRIVATE_INPUTS_ENV, "PAPERCLIP_CHAT_INSTALLED_PLAN", "PAPERCLIP_CHAT_OPERATOR_INSTALL_RECEIPT", "PAPERCLIP_CHAT_INSTALLED_SOURCE_CONFIG"].includes(key)) delete env[key];
if (executions.length) {
const profileNames = new Set(executions.flatMap(execution => [execution.profile.credential, execution.environment.credential].filter(Boolean)));
for (const name of CREDENTIAL_NAMES) if (!profileNames.has(name)) delete env[name];
}
return env;
}
/** Public fingerprint projection contains reference names and field names only. */
export function chatPrivateInputNameIdentity(executionId: string, env: NodeJS.ProcessEnv = process.env) {
const reference = env[CHAT_PRIVATE_INPUTS_ENV], root = env.PAPERCLIP_RUNNER_E2E_TEMP_ROOT;
if (!reference) return null;
if (!root) throw new Error("Selected private input ownership root is missing");
const envelope = readPrivateJson(reference, root) as PrivateEnvelope;
if (envelope.executionId !== executionId) throw new Error("Private input identity belongs to another execution");
const external = envelope.external as { providers?: Record<string, { prerequisites?: Record<string, unknown> }> } | undefined;
return { ...(envelope.oracleExpectation ? { oracleExpectation: envelope.oracleExpectation } : {}), credentialNames: envelope.declaredCredentialNames ?? Object.keys(envelope.credentials).sort(), referenceNames: envelope.declaredReferenceNames ?? Object.keys(envelope.references).sort(),
externalRequirements: envelope.declaredExternalRequirements ?? Object.entries(external?.providers ?? {}).map(([provider, binding]) => ({ provider, fields: Object.keys(binding.prerequisites ?? {}).sort() })) };
}
/** Profile/environment credentials remain genuine launch prerequisites; selected
* case-only inputs are evaluated within its reserved setup, never batch admission. */
export function chatLaunchCredentialNames(execution: MatrixExecution, descriptor?: PrivateInputDescriptor): CredentialName[] {
const selected = new Set(descriptor?.requiredCredentials ?? []);
const launch = new Set([execution.profile.credential, execution.environment.credential].filter(Boolean));
return execution.requiredCredentials.filter(name => !selected.has(name) || launch.has(name));
}
export function chatPrivateSetupBlockers(inputs: ChatPrivateInputs | undefined, descriptor?: PrivateInputDescriptor): string[] {
const requirements = chatPrivateInputRequirements(descriptor); const result = [...inputs?.setupBlockers ?? []];
for (const name of requirements.credentialNames) if (!inputs?.credentials[name]?.trim()) result.push(`missing-credential:${name}`);
if (requirements.external && !inputs?.configuration("external-provider")) result.push(`unavailable-external-configuration:${requirements.external.provider}`);
return [...new Set(result)].sort();
}
export function verifyRetainedChatOracleIdentity(identity: ReturnType<typeof chatPrivateInputNameIdentity> | undefined, descriptor: PrivateInputDescriptor, env: NodeJS.ProcessEnv = process.env) {
const requirement = descriptor.externalInput; if (!requirement || !oracleSemanticFields(requirement.provider, requirement.fields).length) return;
const filename = env[externalConfigEnv]; if (!filename) throw new Error("Retained private oracle expectations unavailable; evidence cannot qualify");
const config = readPrivateJson(filename) as { schemaVersion?: number; providers?: Record<string, { prerequisites?: Record<string, string> }> };
if (config.schemaVersion !== 1) throw new Error("Retained private oracle configuration unavailable");
verifyRetainedPrivateOracle(identity?.oracleExpectation, requirement.provider, requirement.fields, config.providers?.[requirement.provider]?.prerequisites ?? {}, env);
}