Files
PaperClipAI/tests/runner-e2e/private-diagnostics.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

114 lines
8.3 KiB
TypeScript

import { lstatSync, readFileSync, realpathSync, renameSync, writeFileSync } from "node:fs";
import path from "node:path";
import { StringDecoder } from "node:string_decoder";
import { findSecretLeak, normalizedSecrets, redactText } from "./redaction.js";
export const PRIVATE_DIAGNOSTICS_ENV = "PAPERCLIP_RUNNER_E2E_PRIVATE_DIAGNOSTICS";
interface Envelope { version: 1; executionIds: string[]; secrets: string[]; productDelivered?: string[] }
function checkedPath(filename: string, temporaryRoot: string) {
const parent = path.join(temporaryRoot, "selected-private-inputs");
if (filename !== path.join(parent, "diagnostic-redaction.json") || realpathSync(parent) !== parent) throw new Error("Invalid private diagnostic registry path");
const directory = lstatSync(parent); const file = lstatSync(filename);
if (!directory.isDirectory() || directory.mode & 0o077 || file.isSymbolicLink() || !file.isFile() || file.mode & 0o077 || file.size > 4 * 1024 * 1024
|| typeof process.getuid === "function" && (directory.uid !== process.getuid() || file.uid !== process.getuid())) throw new Error("Private diagnostic registry ownership or permissions are invalid");
}
function read(filename: string, temporaryRoot: string): Envelope {
checkedPath(filename, temporaryRoot);
const value = JSON.parse(readFileSync(filename, "utf8")) as Envelope;
if (value.version !== 1 || !Array.isArray(value.executionIds) || !value.executionIds.every(id => typeof id === "string") || !Array.isArray(value.secrets) || !value.secrets.every(secret => typeof secret === "string")
|| value.productDelivered !== undefined && (!Array.isArray(value.productDelivered) || !value.productDelivered.every(secret => typeof secret === "string"))) throw new Error("Invalid private diagnostic registry");
return value;
}
export function stagePrivateDiagnostics(temporaryRoot: string, executionIds: string[], secrets: readonly string[]) {
const filename = path.join(temporaryRoot, "selected-private-inputs", "diagnostic-redaction.json");
writeFileSync(filename, JSON.stringify({ version: 1, executionIds, secrets: normalizedSecrets(secrets) } satisfies Envelope), { mode: 0o600, flag: "wx" });
checkedPath(filename, temporaryRoot); return filename;
}
export function readPrivateDiagnosticSecrets(filename: string, temporaryRoot: string) { return read(filename, temporaryRoot).secrets; }
export function readProductDeliveredSecrets(filename: string, temporaryRoot: string) { return read(filename, temporaryRoot).productDelivered ?? []; }
const productDelivered = new Set<string>();
/** A credential the product itself hands the person, such as the invitation
* join link in the agent's chat reply. It stays a private secret: every
* published artifact redacts it, and logs, run logs and the workspace must not
* hold it. Only the product's embedded database may store it, because the
* reply the person reads is saved there behind the app's access checks. */
export function registerProductDeliveredSecret(secrets: string[], value: string, env: NodeJS.ProcessEnv = process.env) {
if (!value) throw new Error("Product-delivered secret must be a non-empty string");
productDelivered.add(value);
const filename = env[PRIVATE_DIAGNOSTICS_ENV], temporaryRoot = env.PAPERCLIP_RUNNER_E2E_TEMP_ROOT;
if (filename && temporaryRoot) {
const current = read(filename, temporaryRoot);
const temporary = `${filename}.${process.pid}.delivered`;
writeFileSync(temporary, JSON.stringify({ ...current, productDelivered: [...new Set([...(current.productDelivered ?? []), value])] }), { mode: 0o600, flag: "wx" });
renameSync(temporary, filename);
}
secrets.push(value);
}
/** True for a value the product legitimately returns in its own API payloads;
* published evidence still redacts it. */
export function isProductDeliveredSecret(value: string) { return productDelivered.has(value); }
/** Existing case code continues to push into its private redaction sink. Each
* registration is durable before the subsequent instrumented operation starts;
* only the launcher/test host receives this private path, never the product. */
export function trackedPrivateSecrets(initial: string[], executionId: string, env: NodeJS.ProcessEnv = process.env): string[] {
const filename = env[PRIVATE_DIAGNOSTICS_ENV]; if (!filename) return initial;
const temporaryRoot = env.PAPERCLIP_RUNNER_E2E_TEMP_ROOT;
if (!temporaryRoot) throw new Error("Private diagnostic registry requires an owned temporary root");
const previous = read(filename, temporaryRoot);
if (!previous.executionIds.includes(executionId)) throw new Error("Private diagnostic registry execution mismatch");
const persist = (values: string[]) => {
const current = read(filename, temporaryRoot);
const temporary = `${filename}.${process.pid}.next`;
writeFileSync(temporary, JSON.stringify({ ...current, secrets: normalizedSecrets([...current.secrets, ...values]) }), { mode: 0o600, flag: "wx" });
renameSync(temporary, filename);
};
persist(initial);
return new Proxy(initial, { set(target, property, value) {
if (property !== "length" && typeof value !== "string") throw new Error("Private diagnostic secret registration must be a string");
const next = [...target]; Reflect.set(next, property, value); persist(next);
return Reflect.set(target, property, value);
} });
}
export interface DiagnosticSecretExpansion {
readonly kind: "expanded-private-diagnostic-secrets";
readonly values: readonly string[];
}
// Provenance is process-local. Never infer raw-root status from credential bytes:
// independently registered roots can legitimately equal another root's encoding.
const rawDiagnosticRoots = new WeakMap<DiagnosticSecretExpansion, readonly string[]>();
export function diagnosticSecrets(input: readonly string[] | DiagnosticSecretExpansion): DiagnosticSecretExpansion {
if (!Array.isArray(input)) {
const expanded = input as DiagnosticSecretExpansion;
if (!rawDiagnosticRoots.has(expanded)) throw new Error("Diagnostic expansion has no raw-root provenance");
return expanded;
}
const roots = normalizedSecrets(input);
const variants = roots.flatMap(value => {
let structured = false; try { const parsed = JSON.parse(value); structured = parsed !== null && typeof parsed === "object"; } catch { /* Ordinary secret. */ }
return [value, JSON.stringify(value).slice(1, -1), encodeURIComponent(value), Buffer.from(value).toString("base64"), Buffer.from(value).toString("base64url"),
value.replace(/&/g, "&amp;").replace(/</g, "&lt;").replace(/>/g, "&gt;").replace(/"/g, "&quot;").replace(/'/g, "&#39;"),
...value.split(/\r?\n/).filter(line => structured ? line.length >= 8 : line.length > 0)];
});
const expanded: DiagnosticSecretExpansion = Object.freeze({ kind: "expanded-private-diagnostic-secrets", values: Object.freeze(normalizedSecrets(variants)) });
rawDiagnosticRoots.set(expanded, Object.freeze(roots));
return expanded;
}
/** No raw byte reaches a console or log. UTF-8 and secret boundaries may span
* chunks. Oversized unterminated lines are withheld whole, with a truthful
* omission marker; arbitrary memory growth never forces an unsafe flush. */
export function createPrivateDiagnosticStream(secrets: () => readonly string[], emit: (safe: string) => void, maximumLine = 1024 * 1024) {
const decoder = new StringDecoder("utf8"); let pending = ""; let suppressing = false; let failed = false; let ended = false; let leaked = false;
const line = (value: string) => {
try { const values = diagnosticSecrets(secrets()).values; leaked ||= !!findSecretLeak(value, values); emit(redactText(value, values)); }
catch { failed = true; emit("[diagnostics withheld: private redaction unavailable]\n"); }
};
const consume = (text: string) => {
for (const fragment of text.match(/[^\n]*\n|[^\n]+$/g) ?? []) {
if (!suppressing) pending += fragment;
if (pending.length > maximumLine) { pending = ""; suppressing = true; emit("[diagnostic line withheld: size bound exceeded]\n"); }
if (fragment.endsWith("\n")) { if (!suppressing) line(pending); pending = ""; suppressing = false; }
}
};
return { write(chunk: Buffer) { if (!ended) consume(decoder.write(chunk)); }, end() { if (ended) return; ended = true; consume(decoder.end()); if (pending && !suppressing) line(pending); pending = ""; }, failed() { return failed; }, redactedSecret() { return leaked; } };
}