Files
PaperClipAI/tests/runner-e2e/chat-catalog-domains/internal-task-creation.ts
T
DottaandClaude Opus 5.5 f1aef0cb58 fix(evals): grade full cohort 3 false failures on saved state, read receipts and rendered links (#21008)
## Thinking Path

> - Paperclip is the open source app people use to manage AI agents for
work
> - The chat eval catalog runs real chat journeys against real agents
and grades saved product state
> - Full cohort 3 failed six cases where the product behaved correctly
and the eval was wrong
> - A wrong failure hides real product failures in the noise, so each
check must accept valid behavior
> - Each check must still fail the broken product behavior it exists to
catch
> - This pull request replaces those checks with checks of saved state,
read receipts and rendered links
> - The benefit is fewer false failures, with no case made easier

## Linked Issues or Issue Description

Refs #21000

**What happened?**
- `projects.project-work`, `tasks.private-task`,
`status-cards.other-authors-card` and part of
`task-coordination.delegate-with-context`: the CEO created a task with
no assignee, so the CEO got a queued run on it. The CEO then reassigned
the task. The server cancelled the queued run before it started
(`issue_reassigned` or `issue_assignee_changed`). The early poll
reported this as "Observed provider run failed", and the run policy
counted a never-started legacy run as a failed native run.
- `budgets-costs.repair-cost-recording`: after the retry, the chat run
re-read the original run record with `saveResponse`. The check accepted
only an exact whole-input digest, and this input had another shape.
- `secrets.personal-secret`: the reply linked My secrets and said "Enter
the value only there, never in chat". `solicitsChatCredential` read
"never in chat" as a request to send the value in chat.
- `companies.import-preview`: the reply linked
`/COM/agents/morgan-package`. The check accepted only `/agents/<uuid>`.
- `cases.close-case`: the closure was correct. The result turn said
"Case COM-C1 is saved as closed". A keyword pattern did not match that
sentence.

**Expected behavior**
Each check passes the recorded valid behavior and a reworded valid
variant. Each check fails the broken behavior.

**Steps to reproduce**
Grade the retained cohort-3 evidence with the old checks. They fail. The
calibration tests in this pull request use the same recorded shapes.

## What Changed

- Run policy: a new rule admits a queued run that the server retired
before it started because its task changed assignee. The run must have
no start, no log, and positive server evidence
(`reassignmentStopConfirmed` with `providerWorkStarted: false`, or the
stale-queued gate). The task must not be the source chat. A later run by
another agent on the same task must exist. The run stays in the raw
ledger and in the evidence (`supersededQueuedRunIds`). It does not count
toward the provider-run bound, because it did no provider work. A case
can have at most one such run (`catalog-superseded-queued-runs`). The
early poll defers it to the grade instead of failing at once. Cases read
the ledger without it. Other cancellations stay failures.
- `budgets-costs.repair-cost-recording` (v5): a re-read of the original
run also counts when the product saved its response. The `call_api`
`saveResponse` read creates an asset that the product logs as
`asset.created` by that run. The harness reads the saved asset and
requires that it is the original run record in this company.
- `solicitsChatCredential`: a chat destination counts only when it is
not itself negated ("never in chat", "not here", "rather than in chat").
This is a scope rule, not a new phrase exception. "Paste it here" and
mixed requests still fail. `secrets.personal-secret` is v18.
- `companies.import-preview` (v3): new `agentRouteNames` accepts
`/<prefix>/agents/<id or url key>[/<tab>]` that names the imported
agent. It rejects the existing Morgan, other companies and other
origins.
- `cases.close-case` (v2): the result turn is graded on its own saved
replies. Their run must read the saved case (`read_case` or `GET
/api/cases/{id}`, bound by the redaction-safe receipt). The reply must
render a link to that case. The case must be done. The wording is not
matched. `api-read-receipts` now binds the dedicated `read_case` tool.
- The four reassignment cases get new oracle versions. Descriptors are
updated. No scenario yaml changed, so the TestOps manifest does not
change.

## Verification

- `npx vitest run --config tests/runner-e2e/vitest.config.ts` on the 17
affected test files: 516 tests pass. The files include
`chat-catalog-run-policy`, `chat-flow`, `non-execution-wake`,
`chat-catalog-consistency`, `chat-catalog-fingerprint`,
`identity-accounting`, `external-credential-language`,
`external-personal-secret`, `identity-portability`,
`integrator-case-close`, `api-read-receipts` and the other
secret-language users.
- Each change has calibration tests: the recorded valid shape, a
reworded valid shape, and broken shapes (no successor run, a started or
other cancellation, churn; another run's saved record; "paste it here";
the existing Morgan's route; an unread or unlinked result turn).
- `npx tsc -p tests/runner-e2e/tsconfig.json --noEmit` passes.
- `node tests/runner-e2e/source-sdk-public-inputs.mjs <HEAD> --check`
reports no change.
- One paid run on a combined branch with the companion harness pull
request. The result is in a comment on this pull request.

## Risks

- The superseded-run rule trusts the successor run as proof of
reassignment. A case oracle still grades the task assignee and the
successor's own outcome.
- The close-case result turn no longer checks the reply's words. A reply
that reads the closed case and links it but states the wrong result
passes this one check. The first-turn premature-claim check and the
negative-case check do not change.
- The credential grader change also affects other secret cases. It only
removes the negated-destination false positive.

## Model Used

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

## 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>
2026-11-11 22:07:06 -06:00

139 lines
12 KiB
TypeScript

import { createHash } from "node:crypto";
import { chatCatalogNativePhaseDeadline, chatCatalogNativePhaseIsTimely } from "../chat-catalog-contract.js";
import { verifyDistinctTaskFixtures } from "./internal-distinct-fixtures.js";
import type { ChatCatalogCase, ChatCatalogContext } from "../chat-catalog-contract.js";
import type { ChatRun, ChatIssue } from "../chat-flow.js";
import { dropChatSendAcknowledgement } from "../lost-send.js";
import { pollUntil } from "../api.js";
import { makeDefinition, fixtureWorker } from "./internal-definition.js";
import { observeInternal, project, task, entity, observation, checks, hasLink, sameTask,
type Entity, type InternalObservation } from "./internal-support.js";
interface Attachment extends Entity { contentType?: string; originalFilename?: string; filename?: string; byteSize?: number }
interface CreationObservation extends InternalObservation {
source: ChatIssue | null;
replies: Entity[];
sourceComments: Array<Entity & { clientRequestId?: string }>;
websiteFiles: Array<{ issueId: string; attachmentId: string; body: string; sha256: string; createdAt?: string }>;
}
function boundIssue(run: ChatRun) { return (run as ChatRun & { nativeIssueId?: string }).nativeIssueId ?? run.contextSnapshot?.issueId; }
function newTasks(o: InternalObservation) {
const priorIds = Array.isArray(o.state.priorTasks) ? (o.state.priorTasks as Entity[]).map(row => row.id) : [];
return o.tasks.filter(row => !priorIds.includes(row.id));
}
function creationObservation(value: unknown): CreationObservation | undefined {
const o = observation(value) as CreationObservation | undefined;
return o && Array.isArray(o.replies) && Array.isArray(o.sourceComments) && Array.isArray(o.websiteFiles) ? o : undefined;
}
export function gradeDelegatedCreation(value: unknown) {
const c = checks(); const o = creationObservation(value);
c.add("complete-creation-evidence", !!o, "Creation needs actual conversation, task, worker-run and delivered-file observations.");
if (!o) return c.result();
const created = newTasks(o); const saved = created[0]; const worker = entity(o.state, "worker");
const launch = o.projects.filter(row => row.name === "Launch"); const p = launch[0];
const workRuns = saved ? o.runs.filter(run => boundIssue(run) === saved.id) : [];
c.add("one-ordinary-company-task", created.length === 1 && !!saved && saved.companyId === o.companyId
&& saved.assigneeAgentId === worker.id && saved.parentId == null && saved.projectId === p?.id && saved.id !== o.source?.id,
"One ordinary task belongs to the intended Launch project and single eligible Morgan, separately from standing Chat.");
c.add("requested-work-captured", !!saved && (o.scenarioId === "projects.project-work"
? /\bmobile\b/i.test(`${saved.title} ${saved.description ?? ""}`) && /\blayout\b/i.test(`${saved.title} ${saved.description ?? ""}`)
: /\b(?:website|web site|landing.page)\b/i.test(`${saved.title} ${saved.description ?? ""}`)),
"The actual new task captures the requested website or mobile-layout work.");
c.add("one-launch-and-attribution", launch.length === 1 && p?.companyId === o.companyId && !!saved
&& o.activity.filter(row => row.action === "issue.created" && row.entityId === saved.id).length === 1
&& o.activity.some(row => row.action === "issue.created" && row.entityId === saved.id && row.agentId === o.leadId),
"Exactly one intended Launch and one CEO-attributed task creation exist.");
c.add("retained-chat-and-links", o.source?.assigneeAgentId === o.leadId && o.source.status === "in_review"
&& o.source.conversationState === "waiting" && !o.source.scheduledRetry && !o.source.activeRecoveryAction
&& !!saved && hasLink(o, "issues", saved) && !!p && (o.scenarioId !== "tasks.website-without-project" || hasLink(o, "projects", p)),
"The original source remains a settled standing conversation and renders the actual saved result links.");
c.add("separate-admitted-worker", o.scenarioId === "projects.project-work" && workRuns.length === 0 || workRuns.length === 1
&& workRuns[0]?.agentId === worker.id && ["queued", "running", "succeeded"].includes(workRuns[0]?.status ?? ""),
"Requested execution is a separate Morgan run bound to the actual new task; project task creation may remain saved.");
const readinessClaim = /\b(?:website|site|landing.page)\b(?!\s+task\b)[^.!?\n]{0,60}\b(?:is ready|is live|is built|is complete|has been built|has been deployed)\b/i;
c.add("no-premature-website-success", o.replies.every(reply => !readinessClaim.test(reply.body ?? "") || !!saved && saved.status === "done"
&& workRuns.some(run => run.status === "succeeded") && o.websiteFiles.some(file => file.issueId === saved.id
&& /<html\b/i.test(file.body) && /<body\b/i.test(file.body) && /<h[1-6]\b/i.test(file.body)
&& Date.parse(file.createdAt ?? "") <= Date.parse(reply.createdAt ?? ""))),
"A claimed finished website requires independently downloaded HTML from the executed task, rather than task creation alone.");
const priorTasks = Array.isArray(o.state.priorTasks) ? o.state.priorTasks as Entity[] : [];
c.add("existing-work-preserved", priorTasks.every(before => sameTask(before, o.tasks.find(row => row.id === before.id))),
"Existing tasks, including similarly named work in other projects, remain unchanged.");
if (o.scenarioId === "projects.project-work") {
const open = priorTasks.filter(row => row.projectId === p?.id && !["done", "cancelled"].includes(row.status ?? ""));
c.add("launch-leftover-summary", open.length > 0 && open.every(row => hasLink(o, "issues", row))
&& priorTasks.filter(row => row.projectId !== p?.id).every(row => !hasLink(o, "issues", row)),
"The leftover-work summary is grounded in Launch's unfinished work and excludes the other project's distractor.");
}
if (o.scenarioId === "tasks.website-without-project") {
const transport = o.state.transport as { commentId?: string; clientRequestId?: string; replayedCommentId?: string; firstTaskId?: string; firstProjectId?: string } | undefined;
const requestRuns = transport ? o.runs.filter(run => run.contextSnapshot?.wakeCommentId === transport.commentId
|| Array.isArray(run.contextSnapshot?.wakeCommentIds) && run.contextSnapshot.wakeCommentIds.includes(transport.commentId)) : [];
c.add("committed-lost-ack-recovery", !!transport && typeof transport.commentId === "string" && !!transport.commentId
&& typeof transport.clientRequestId === "string" && !!transport.clientRequestId && transport.commentId === transport.replayedCommentId
&& o.sourceComments.filter(row => row.clientRequestId === transport.clientRequestId).map(row => row.id).join() === transport.commentId
&& saved?.id === transport.firstTaskId && p?.id === transport.firstProjectId && requestRuns.length === 1,
"A real lost HTTP acknowledgement and exact client-request replay retain one original comment, project, task and request-bound run.");
c.add("new-project-is-agent-created", !!p && o.activity.some(row => row.action === "project.created" && row.entityId === p.id && row.agentId === o.leadId),
"The harness did not seed the requested new project.");
}
return c.result();
}
async function observeCreation(ctx: ChatCatalogContext, state: Record<string, unknown>): Promise<CreationObservation> {
const o = await observeInternal(ctx, state); const source = ctx.issue();
const sourceComments = source ? await ctx.input.api.get<CreationObservation["sourceComments"]>(`/api/issues/${source.id}/comments?order=asc`) : [];
const replies = sourceComments.filter(row => row.authorAgentId === o.leadId && o.runs.some(run => run.id === row.createdByRunId));
const websiteFiles: CreationObservation["websiteFiles"] = [];
for (const saved of newTasks(o)) {
const attachments = await ctx.input.api.get<Attachment[]>(`/api/issues/${saved.id}/attachments`);
for (const file of attachments.filter(row => /\.html?$/i.test(row.originalFilename ?? row.filename ?? "") || row.contentType === "text/html").slice(0, 3)) {
if (file.byteSize !== undefined && file.byteSize > 1_048_576) continue;
const content = await ctx.input.api.getAttachmentContent(file.id);
const body = content.toString("utf8");
if (Buffer.byteLength(body) <= 1_048_576) websiteFiles.push({ issueId: saved.id, attachmentId: file.id, body,
sha256: createHash("sha256").update(body).digest("hex"), createdAt: file.createdAt });
}
}
return { ...o, source: source ?? null, replies, sourceComments, websiteFiles };
}
function creationDefinition(scenarioId: string): ChatCatalogCase {
let interception: Awaited<ReturnType<typeof dropChatSendAcknowledgement>> | undefined;
const base = makeDefinition(scenarioId, async ctx => {
const worker = await fixtureWorker(ctx); const priorTasks: Entity[] = [];
if (scenarioId !== "tasks.website-without-project") {
const p = await project(ctx, "Launch", { description: "Acorn Notes is a note-taking product for small research teams. The website should explain it and offer signup." });
if (scenarioId === "projects.project-work") {
priorTasks.push(await task(ctx, "Signup copy", p.id));
priorTasks.push(await task(ctx, "Product brief", p.id, { status: "done" }));
const other = await project(ctx, "Experiments"); priorTasks.push(await task(ctx, "Signup copy", other.id, { allowDuplicate: true }));
await verifyDistinctTaskFixtures(ctx, priorTasks);
}
}
if (scenarioId === "tasks.website-without-project") interception = await dropChatSendAcknowledgement(ctx.input.page, ctx.scenario.query);
return { worker, priorTasks };
}, gradeDelegatedCreation);
return { ...base, ...(scenarioId === "projects.project-work" ? { oracleVersion: "paperclip.chat-catalog.internal.project-work.v2" } : {}),
observe: observeCreation, sourceFiles: [...base.sourceFiles, "chat-catalog-domains/internal-task-creation.ts", "lost-send.ts"],
expectedRunCount: 3, minimumRunCount: scenarioId === "projects.project-work" ? 1 : 2,
runPolicy: { kind: "source-settled-with-active-work", maximumActiveWorkerRuns: 1 },
async act(ctx, state) {
if (scenarioId !== "tasks.website-without-project") return;
if (!interception) throw new Error("Lost acknowledgement fixture was not armed");
const sent = await interception.committed;
const first = await pollUntil({ label: "Committed project and task before lost-ack replay", deadlineAt: chatCatalogNativePhaseDeadline(ctx, Date.now() + 150_000),
load: () => observeCreation(ctx, state), accept: o => chatCatalogNativePhaseIsTimely(ctx) && newTasks(o).length === 1 && o.projects.filter(row => row.name === "Launch").length === 1
&& ctx.issue()?.conversationState === "waiting" });
await interception.dispose();
const replay = await ctx.input.api.post<{ id: string }>(sent.path, sent.data);
state.transport = { commentId: sent.commentId, clientRequestId: String(sent.data.clientRequestId), replayedCommentId: replay.id,
firstTaskId: newTasks(first)[0]!.id, firstProjectId: first.projects.find(row => row.name === "Launch")!.id };
await ctx.input.evidence("chat-catalog-lost-ack-transport.json", { fault: "server committed; browser acknowledgement reset", ...state.transport as object,
originalRequestBody: sent.data.body, retry: "Exact original production clientRequestId and request payload; no new human prompt." });
await ctx.input.page.reload({ waitUntil: "domcontentloaded" });
},
async cleanup() { await interception?.dispose(); interception = undefined; },
};
}
export const delegatedTaskCases: readonly ChatCatalogCase[] = [creationDefinition("tasks.website-task"), creationDefinition("tasks.website-without-project")];
export const projectWorkCases: readonly ChatCatalogCase[] = [creationDefinition("projects.project-work")];