Files
PaperClipAI/server/src/__tests__/codex-local-execute.test.ts
T
DottaandPaperclip e26d787928 Shorten continuation prompts and verify question tool guidance (#13574)
## Thinking Path

> - Paperclip is the open source app people use to manage AI agents for
work.
> - Agents must continue tasks using user answers without losing earlier
requirements or approval gates.
> - The wake prompt mixed human decisions with prior tool evidence and
repeated detailed question instructions.
> - Those instructions belong with the question tool, with a short
routing hint in the wake.
> - The Runner evals need to prove that answers, approvals, and
completed work survive later turns.
> - This PR shortens the prompts, separates authenticated answers, and
adds continuation tests with useful screenshots.

## Linked Issues or Issue Description

Refs #13517. This is a follow-up to the merged onboarding skill and
Runner E2E work. Related #13539 covers responses received while a run is
active; this PR preserves its cases and adds continuation coverage.
Existing continuation/recovery and question PRs were searched; none
covers this prompt/documentation and eval change.

**What existing behavior does this improve?**

The instructions sent when an agent continues a task, the native
human-input tool documentation, and the evidence captured by Runner
full-stack E2E.

**Current behavior**

The wake repeats a long question-tool guide. Human answers appear
alongside untrusted prior results. Screenshot capture can finish at DOM
load while the task still shows a spinner, even when backend behavior
checks pass.

**Proposed behavior**

Keep earlier requirements unless the user changes them. Treat
clarification as distinct from approval. Give authenticated human
responses a scoped field. Keep tool and agent results as evidence. Put
detailed question behavior in the tool descriptor and retain one routing
sentence in the native wake. Wait for the correct task and loaded
conversation before taking screenshots.

**Reason and benefit**

Reduce repeated prompt text and make authority boundaries clear. Test
that real question cards, later answers, approval gates, and completed
child tasks still work. Make screenshots useful for human review.

## What Changed

- Shorten shared continuation instructions for legacy and native
runners. Separate authenticated user responses from tool results and
agent summaries.
- Remove the detailed question guide from native wake prompts. Keep its
behavior in the canonical `request_human_input` descriptor and existing
payload schema. Regenerate semantic contracts and fixture hashes.
- Add five continuation cases across four local profiles. Add a
dedicated choice-then-text case for native Codex and native Claude. All
22 cells join the shared full E2E campaign.
- Cover revised scope, clarification without approval, hostile
instructions in a handoff file, and reuse of a completed child after
restart. Keep production instructions and fixed user facts.
- Capture continuation screenshots only when the intended task and
conversation have rendered. Add provider-free browser regressions for
loaders and wrong-task capture.
- Preserve current master’s extra tool and onboarding cases. The default
campaign now contains 166 cells; 35 manual everyday cells remain
separate.

## Verification

- `pnpm -r typecheck`: passed after replay on current master.
- `pnpm test:e2e:runner:unit`: 340 passed. Harness typecheck passed.
- `PAPERCLIP_PLAYWRIGHT_CHANNEL=chrome pnpm
test:e2e:runner:browser-support`: 4 passed. These tests failed against
immediate screenshot capture and passed after the fix.
- Focused continuation and native-input tests: 36 passed locally. The
tool-authority suite could not initialize embedded PostgreSQL locally,
including one isolated retry; its 17 assertions did not run locally. The
full remote server shards passed on this PR commit.
- `pnpm build`: passed after replay on current master. `pnpm test:run`
was attempted locally but hit the same embedded PostgreSQL
initialization failure; the remaining local run was stopped after
complete remote CI passed. This is not claimed as a full local test
pass.
- [Full PR
CI](https://github.com/paperclipai/paperclip/actions/runs/35232755685):
passed on `6a22128c14f4552d0613a6d9a25955db4a1ed02f`. All
server/chat/workspace/serialized shards, browser shards, Runner checks,
typecheck, build, canary and policy checks passed. The isolated native
Runner build and security checks also passed: 57 successful checks, with
two expected Storybook skips.
- Greptile reviewed the exact PR head at 5/5, with no findings or
unresolved review threads. The PR has no merge conflicts.
- [Live question-docs
report](https://pages.paperclip.ing/runner-e2e-question-docs-35227647794/):
3/3 passed at source `83dd132f2` before replay on master. Native Codex
and Claude each asked a choice, waited, asked a text question, and saved
both answers. Claude also passed a completed-child restart case. All
three native turns are checked for absence of the old question block.
- [Earlier continuation
report](https://pages.paperclip.ing/runner-e2e-continuation-35154943615/):
all five continuation cases passed on native Claude. The report retains
campaign and revision provenance and separately shows two unresolved
onboarding behavior failures.
- [Before/after prompt
report](https://pages.paperclip.ing/runner-prompt-comparison-20260917/):
full text, current recorded Claude inputs, and reproducible
reference-token counts. The controlled wake comparison removes 401
reference tokens; the net counted input reduction is 339 after charging
the larger tool description. These are text-size estimates, not measured
billing savings.

## Risks

- Prompt wording affects model behavior. Live results cover the stated
cases, not every provider or conversation. Legacy profiles are
registered but were not rerun for this change.
- The optional continuation field changes prompt data only; there is no
database migration or new production API.
- Authenticated answer projection excludes generated summaries and
agent-resolved interactions. It preserves the answer’s question or
approval scope.
- The screenshot guard can expose UI loading failures that earlier runs
hid. Backend grading alone no longer makes those captures valid.
- The two prior onboarding failures remain separate product issues: work
before acceptance and a missing saved plan. This PR does not claim the
entire onboarding suite passes.

## Model Used

OpenAI Codex, GPT-6, with reasoning, repository tools, code execution,
and browser verification. The exact deployed model identifier and
context-window size are not exposed in this session.

## 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 — targeted tests above; the
full local database-startup limit is documented
- [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
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge

Co-authored-by: Paperclip <noreply@paperclip.ing>
2026-09-17 09:32:54 -05:00

1648 lines
65 KiB
TypeScript

import { afterEach, describe, expect, it, vi } from "vitest";
import fs from "node:fs/promises";
import os from "node:os";
import path from "node:path";
import { runChildProcess } from "@paperclipai/adapter-utils/server-utils";
import { execute } from "@paperclipai/adapter-codex-local/server";
import { buildPaperclipTaskMarkdown } from "../services/heartbeat.js";
import { AGENT_CHAT_DIRECTIVE } from "../services/agent-conversations.js";
async function writeFakeCodexCommand(commandPath: string): Promise<void> {
const script = `#!/usr/bin/env node
const fs = require("node:fs");
const capturePath = process.env.PAPERCLIP_TEST_CAPTURE_PATH;
const payload = {
argv: process.argv.slice(2),
prompt: fs.readFileSync(0, "utf8"),
codexHome: process.env.CODEX_HOME || null,
codexConfigContents: process.env.CODEX_HOME && fs.existsSync(process.env.CODEX_HOME + "/config.toml")
? fs.readFileSync(process.env.CODEX_HOME + "/config.toml", "utf8")
: null,
paperclipWakePayloadJson: process.env.PAPERCLIP_WAKE_PAYLOAD_JSON || null,
paperclipApiUrl: process.env.PAPERCLIP_API_URL || null,
paperclipApiKey: process.env.PAPERCLIP_API_KEY || null,
paperclipApiBridgeMode: process.env.PAPERCLIP_API_BRIDGE_MODE || null,
paperclipEnvKeys: Object.keys(process.env)
.filter((key) => key.startsWith("PAPERCLIP_"))
.sort(),
};
if (capturePath) {
fs.writeFileSync(capturePath, JSON.stringify(payload), "utf8");
}
console.log(JSON.stringify({ type: "thread.started", thread_id: "codex-session-1" }));
console.log(JSON.stringify({ type: "item.completed", item: { type: "agent_message", text: "hello" } }));
console.log(JSON.stringify({ type: "turn.completed", usage: { input_tokens: 1, cached_input_tokens: 0, output_tokens: 1 } }));
`;
await fs.writeFile(commandPath, script, "utf8");
await fs.chmod(commandPath, 0o755);
}
async function writeFailingCodexCommand(commandPath: string, errorMessage: string): Promise<void> {
const script = `#!/usr/bin/env node
console.log(JSON.stringify({ type: "error", message: ${JSON.stringify(errorMessage)} }));
process.exit(1);
`;
await fs.writeFile(commandPath, script, "utf8");
await fs.chmod(commandPath, 0o755);
}
type CapturePayload = {
argv: string[];
prompt: string;
codexHome: string | null;
codexConfigContents?: string | null;
paperclipWakePayloadJson: string | null;
paperclipApiUrl?: string | null;
paperclipApiKey?: string | null;
paperclipApiBridgeMode?: string | null;
paperclipEnvKeys: string[];
};
type LogEntry = {
stream: "stdout" | "stderr";
chunk: string;
};
const fakeCodexAuthJson = JSON.stringify({ OPENAI_API_KEY: "sk-test-codex-local" });
const codexHomeOverrides: Array<string | undefined> = [];
afterEach(() => {
while (codexHomeOverrides.length > 0) {
const previous = codexHomeOverrides.pop();
if (previous === undefined) delete process.env.CODEX_HOME;
else process.env.CODEX_HOME = previous;
}
});
async function seedSharedCodexAuth(homeRoot: string): Promise<void> {
const sharedCodexHome = path.join(homeRoot, ".codex");
codexHomeOverrides.push(process.env.CODEX_HOME);
process.env.CODEX_HOME = sharedCodexHome;
await fs.mkdir(sharedCodexHome, { recursive: true });
await fs.writeFile(path.join(sharedCodexHome, "auth.json"), `${fakeCodexAuthJson}\n`, "utf8");
}
function createLocalSandboxRunner() {
let counter = 0;
return {
execute: async (input: {
command: string;
args?: string[];
cwd?: string;
env?: Record<string, string>;
stdin?: string;
timeoutMs?: number;
onLog?: (stream: "stdout" | "stderr", chunk: string) => Promise<void>;
onSpawn?: (meta: { pid: number; startedAt: string }) => Promise<void>;
}) => {
counter += 1;
return runChildProcess(
`sandbox-run-${counter}`,
input.command,
input.args ?? [],
{
cwd: input.cwd ?? process.cwd(),
env: input.env ?? {},
stdin: input.stdin,
timeoutSec: Math.max(1, Math.ceil((input.timeoutMs ?? 30_000) / 1000)),
graceSec: 5,
onLog: input.onLog ?? (async () => {}),
onSpawn: input.onSpawn
? async (meta) => input.onSpawn?.({ pid: meta.pid, startedAt: meta.startedAt })
: undefined,
},
);
},
};
}
describe("codex execute", () => {
it("uses a Paperclip-managed CODEX_HOME outside worktree mode while preserving shared auth and config", async () => {
const root = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-codex-execute-default-"));
const workspace = path.join(root, "workspace");
const commandPath = path.join(root, "codex");
const capturePath = path.join(root, "capture.json");
const sharedCodexHome = path.join(root, "shared-codex-home");
const paperclipHome = path.join(root, "paperclip-home");
const managedCodexHome = path.join(
paperclipHome,
"instances",
"default",
"companies",
"company-1",
"codex-home",
);
await fs.mkdir(workspace, { recursive: true });
await fs.mkdir(sharedCodexHome, { recursive: true });
await fs.writeFile(path.join(sharedCodexHome, "auth.json"), `${fakeCodexAuthJson}\n`, "utf8");
await fs.writeFile(path.join(sharedCodexHome, "config.toml"), 'model = "codex-mini-latest"\n', "utf8");
await writeFakeCodexCommand(commandPath);
const previousHome = process.env.HOME;
const previousPaperclipHome = process.env.PAPERCLIP_HOME;
const previousPaperclipInstanceId = process.env.PAPERCLIP_INSTANCE_ID;
const previousPaperclipInWorktree = process.env.PAPERCLIP_IN_WORKTREE;
const previousCodexHome = process.env.CODEX_HOME;
process.env.HOME = root;
process.env.PAPERCLIP_HOME = paperclipHome;
delete process.env.PAPERCLIP_INSTANCE_ID;
delete process.env.PAPERCLIP_IN_WORKTREE;
process.env.CODEX_HOME = sharedCodexHome;
try {
const logs: LogEntry[] = [];
const result = await execute({
runId: "run-default",
agent: {
id: "agent-1",
companyId: "company-1",
name: "Codex Coder",
adapterType: "codex_local",
adapterConfig: { engine: "cli" },
},
runtime: {
sessionId: null,
sessionParams: null,
sessionDisplayId: null,
taskKey: null,
},
config: {
engine: "cli",
command: commandPath,
cwd: workspace,
env: {
PAPERCLIP_TEST_CAPTURE_PATH: capturePath,
},
promptTemplate: "Follow the paperclip heartbeat.",
},
context: {},
authToken: "run-jwt-token",
onLog: async (stream, chunk) => {
logs.push({ stream, chunk });
},
});
expect(result.exitCode).toBe(0);
expect(result.errorMessage).toBeNull();
expect(result.usage).toEqual({ inputTokens: 1, cachedInputTokens: 0, outputTokens: 1 });
expect(result.usageBasis).toBe("per_run");
expect(result.costUsd).toBeNull();
const capture = JSON.parse(await fs.readFile(capturePath, "utf8")) as CapturePayload;
expect(capture.codexHome).toBe(managedCodexHome);
const managedAuth = path.join(managedCodexHome, "auth.json");
const managedConfig = path.join(managedCodexHome, "config.toml");
expect((await fs.lstat(managedAuth)).isSymbolicLink()).toBe(true);
expect(await fs.realpath(managedAuth)).toBe(await fs.realpath(path.join(sharedCodexHome, "auth.json")));
expect((await fs.lstat(managedConfig)).isFile()).toBe(true);
expect(await fs.readFile(managedConfig, "utf8")).toBe('model = "codex-mini-latest"\n');
await expect(fs.lstat(path.join(sharedCodexHome, "companies", "company-1"))).rejects.toThrow();
expect(logs).toContainEqual(
expect.objectContaining({
stream: "stdout",
chunk: expect.stringContaining("Using Paperclip-managed Codex home"),
}),
);
} finally {
if (previousHome === undefined) delete process.env.HOME;
else process.env.HOME = previousHome;
if (previousPaperclipHome === undefined) delete process.env.PAPERCLIP_HOME;
else process.env.PAPERCLIP_HOME = previousPaperclipHome;
if (previousPaperclipInstanceId === undefined) delete process.env.PAPERCLIP_INSTANCE_ID;
else process.env.PAPERCLIP_INSTANCE_ID = previousPaperclipInstanceId;
if (previousPaperclipInWorktree === undefined) delete process.env.PAPERCLIP_IN_WORKTREE;
else process.env.PAPERCLIP_IN_WORKTREE = previousPaperclipInWorktree;
if (previousCodexHome === undefined) delete process.env.CODEX_HOME;
else process.env.CODEX_HOME = previousCodexHome;
await fs.rm(root, { recursive: true, force: true });
}
});
it("writes managed MCP gateways into Codex config and warns on overlapping direct entries without logging tokens", async () => {
const root = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-codex-managed-mcp-"));
const workspace = path.join(root, "workspace");
const commandPath = path.join(root, "codex");
const capturePath = path.join(root, "capture.json");
const sharedCodexHome = path.join(root, "shared-codex-home");
const paperclipHome = path.join(root, "paperclip-home");
const managedCodexHome = path.join(
paperclipHome,
"instances",
"default",
"companies",
"company-1",
"codex-home",
);
await fs.mkdir(workspace, { recursive: true });
await fs.mkdir(sharedCodexHome, { recursive: true });
await fs.writeFile(path.join(sharedCodexHome, "auth.json"), `${fakeCodexAuthJson}\n`, "utf8");
await fs.writeFile(
path.join(sharedCodexHome, "config.toml"),
[
'model = "codex-mini-latest"',
"",
'[mcp_servers.github]',
'url = "https://raw.example/mcp"',
"",
].join("\n"),
"utf8",
);
await writeFakeCodexCommand(commandPath);
const previousHome = process.env.HOME;
const previousPaperclipHome = process.env.PAPERCLIP_HOME;
const previousPaperclipApiUrl = process.env.PAPERCLIP_API_URL;
const previousPaperclipRuntimeApiUrl = process.env.PAPERCLIP_RUNTIME_API_URL;
const previousCodexHome = process.env.CODEX_HOME;
process.env.HOME = root;
process.env.PAPERCLIP_HOME = paperclipHome;
process.env.PAPERCLIP_API_URL = "http://paperclip.local:3100";
process.env.PAPERCLIP_RUNTIME_API_URL = "http://paperclip.local:3100";
process.env.CODEX_HOME = sharedCodexHome;
try {
const logs: LogEntry[] = [];
const result = await execute({
runId: "run-managed-mcp",
agent: {
id: "agent-1",
companyId: "company-1",
name: "Codex Coder",
adapterType: "codex_local",
adapterConfig: {},
},
runtime: {
sessionId: null,
sessionParams: null,
sessionDisplayId: null,
taskKey: null,
},
config: {
engine: "cli",
command: commandPath,
cwd: workspace,
env: {
PAPERCLIP_TEST_CAPTURE_PATH: capturePath,
},
promptTemplate: "Follow the paperclip heartbeat.",
},
runtimeMcp: {
getServers: () => [
{
name: "github",
url: "http://paperclip.local:3100/api/tool-gateway/gateways/gateway-1/mcp",
token: "pcgw_secret-managed-token",
connectionId: "connection-github",
},
],
},
context: {},
authToken: "run-jwt-token",
onLog: async (stream, chunk) => {
logs.push({ stream, chunk });
},
});
expect(result.exitCode).toBe(0);
expect(result.errorMessage).toBeNull();
const capture = JSON.parse(await fs.readFile(capturePath, "utf8")) as CapturePayload;
const configText = capture.codexConfigContents ?? "";
expect(configText).toContain("[mcp_servers.github]");
expect(configText).toContain("[mcp_servers.\"paperclip-github\"]");
expect(configText).toContain('url = "http://paperclip.local:3100/api/tool-gateway/gateways/gateway-1/mcp"');
expect(configText).toContain('Authorization = "Bearer pcgw_secret-managed-token"');
expect(logs).toEqual(
expect.arrayContaining([
expect.objectContaining({
stream: "stderr",
chunk: expect.stringContaining("Paperclip cannot enforce policies for that direct entry"),
}),
]),
);
expect(JSON.stringify(logs)).not.toContain("pcgw_secret-managed-token");
} finally {
if (previousHome === undefined) delete process.env.HOME;
else process.env.HOME = previousHome;
if (previousPaperclipHome === undefined) delete process.env.PAPERCLIP_HOME;
else process.env.PAPERCLIP_HOME = previousPaperclipHome;
if (previousPaperclipApiUrl === undefined) delete process.env.PAPERCLIP_API_URL;
else process.env.PAPERCLIP_API_URL = previousPaperclipApiUrl;
if (previousPaperclipRuntimeApiUrl === undefined) delete process.env.PAPERCLIP_RUNTIME_API_URL;
else process.env.PAPERCLIP_RUNTIME_API_URL = previousPaperclipRuntimeApiUrl;
if (previousCodexHome === undefined) delete process.env.CODEX_HOME;
else process.env.CODEX_HOME = previousCodexHome;
await fs.rm(root, { recursive: true, force: true });
}
});
it("emits a command note that Codex auto-applies repo-scoped AGENTS.md files", async () => {
const root = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-codex-execute-notes-"));
const workspace = path.join(root, "workspace");
const commandPath = path.join(root, "codex");
const capturePath = path.join(root, "capture.json");
await fs.mkdir(workspace, { recursive: true });
await writeFakeCodexCommand(commandPath);
const previousHome = process.env.HOME;
process.env.HOME = root;
await seedSharedCodexAuth(root);
let commandNotes: string[] = [];
try {
const result = await execute({
runId: "run-notes",
agent: {
id: "agent-1",
companyId: "company-1",
name: "Codex Coder",
adapterType: "codex_local",
adapterConfig: { engine: "cli" },
},
runtime: {
sessionId: null,
sessionParams: null,
sessionDisplayId: null,
taskKey: null,
},
config: {
engine: "cli",
command: commandPath,
cwd: workspace,
env: {
PAPERCLIP_TEST_CAPTURE_PATH: capturePath,
},
promptTemplate: "Follow the paperclip heartbeat.",
},
context: {},
authToken: "run-jwt-token",
onLog: async () => {},
onMeta: async (meta) => {
commandNotes = Array.isArray(meta.commandNotes) ? meta.commandNotes : [];
},
});
expect(result.exitCode).toBe(0);
expect(result.errorMessage).toBeNull();
expect(commandNotes).toContain(
"Codex exec automatically applies repo-scoped AGENTS.md instructions from the current workspace; Paperclip does not currently suppress that discovery.",
);
} finally {
if (previousHome === undefined) delete process.env.HOME;
else process.env.HOME = previousHome;
await fs.rm(root, { recursive: true, force: true });
}
});
it("logs HOME and the resolved executable path in invocation metadata", async () => {
const root = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-codex-execute-meta-"));
const workspace = path.join(root, "workspace");
const binDir = path.join(root, "bin");
const commandPath = path.join(binDir, "codex");
const capturePath = path.join(root, "capture.json");
await fs.mkdir(workspace, { recursive: true });
await fs.mkdir(binDir, { recursive: true });
await writeFakeCodexCommand(commandPath);
const previousHome = process.env.HOME;
const previousPath = process.env.PATH;
process.env.HOME = root;
process.env.PATH = `${binDir}${path.delimiter}${process.env.PATH ?? ""}`;
await seedSharedCodexAuth(root);
let loggedCommand: string | null = null;
let loggedEnv: Record<string, string> = {};
try {
const result = await execute({
runId: "run-meta",
agent: {
id: "agent-1",
companyId: "company-1",
name: "Codex Coder",
adapterType: "codex_local",
adapterConfig: { engine: "cli" },
},
runtime: {
sessionId: null,
sessionParams: null,
sessionDisplayId: null,
taskKey: null,
},
config: {
engine: "cli",
command: "codex",
cwd: workspace,
env: {
PAPERCLIP_TEST_CAPTURE_PATH: capturePath,
},
promptTemplate: "Follow the paperclip heartbeat.",
},
context: {},
authToken: "run-jwt-token",
onLog: async () => {},
onMeta: async (meta) => {
loggedCommand = meta.command;
loggedEnv = meta.env ?? {};
},
});
expect(result.exitCode).toBe(0);
expect(result.errorMessage).toBeNull();
expect(loggedCommand).toBe(commandPath);
expect(loggedEnv.HOME).toBe(root);
expect(loggedEnv.PAPERCLIP_RESOLVED_COMMAND).toBe(commandPath);
} finally {
if (previousHome === undefined) delete process.env.HOME;
else process.env.HOME = previousHome;
if (previousPath === undefined) delete process.env.PATH;
else process.env.PATH = previousPath;
await fs.rm(root, { recursive: true, force: true });
}
});
it("injects bridge env into sandbox-managed remote runs", async () => {
const root = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-codex-execute-sandbox-"));
const localWorkspace = path.join(root, "workspace");
const remoteWorkspace = path.join(root, "sandbox");
const binDir = path.join(root, "bin");
const commandPath = path.join(binDir, "codex");
const capturePath = path.join(remoteWorkspace, "capture.json");
const previousHome = process.env.HOME;
const previousPath = process.env.PATH;
await fs.mkdir(localWorkspace, { recursive: true });
await fs.mkdir(remoteWorkspace, { recursive: true });
await fs.mkdir(binDir, { recursive: true });
await writeFakeCodexCommand(commandPath);
process.env.HOME = root;
process.env.PATH = `${binDir}${path.delimiter}${process.env.PATH ?? ""}`;
await seedSharedCodexAuth(root);
try {
const result = await execute({
runId: "run-sandbox-auth",
agent: {
id: "agent-1",
companyId: "company-1",
name: "Codex Coder",
adapterType: "codex_local",
adapterConfig: { engine: "cli" },
},
runtime: {
sessionId: null,
sessionParams: null,
sessionDisplayId: null,
taskKey: null,
},
config: {
engine: "cli",
command: commandPath,
cwd: localWorkspace,
env: {
PAPERCLIP_TEST_CAPTURE_PATH: capturePath,
},
promptTemplate: "Follow the paperclip heartbeat.",
},
context: {},
executionTarget: {
kind: "remote",
transport: "sandbox",
providerKey: "e2b",
environmentId: "env-1",
leaseId: "lease-1",
remoteCwd: remoteWorkspace,
timeoutMs: 30_000,
runner: createLocalSandboxRunner(),
},
authToken: "run-jwt-token",
onLog: async () => {},
});
expect(result.exitCode).toBe(0);
expect(result.errorMessage).toBeNull();
const capture = JSON.parse(await fs.readFile(capturePath, "utf8")) as CapturePayload;
expect(capture.codexHome).toBe(path.join(remoteWorkspace, ".paperclip-runtime", "codex", "home"));
expect(capture.paperclipApiUrl).toMatch(/^http:\/\/127\.0\.0\.1:\d+$/);
expect(capture.paperclipApiKey).not.toBe("run-jwt-token");
expect(capture.paperclipApiBridgeMode).toBe("queue_v1");
} finally {
if (previousHome === undefined) delete process.env.HOME;
else process.env.HOME = previousHome;
if (previousPath === undefined) delete process.env.PATH;
else process.env.PATH = previousPath;
await fs.rm(root, { recursive: true, force: true });
}
});
it("injects structured Paperclip wake payloads into env and prompt", async () => {
const root = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-codex-execute-wake-"));
const workspace = path.join(root, "workspace");
const commandPath = path.join(root, "codex");
const capturePath = path.join(root, "capture.json");
await fs.mkdir(workspace, { recursive: true });
await writeFakeCodexCommand(commandPath);
const previousHome = process.env.HOME;
process.env.HOME = root;
await seedSharedCodexAuth(root);
try {
const result = await execute({
runId: "run-wake",
agent: {
id: "agent-1",
companyId: "company-1",
name: "Codex Coder",
adapterType: "codex_local",
adapterConfig: { engine: "cli" },
},
runtime: {
sessionId: null,
sessionParams: null,
sessionDisplayId: null,
taskKey: null,
},
config: {
engine: "cli",
command: commandPath,
cwd: workspace,
env: {
PAPERCLIP_TEST_CAPTURE_PATH: capturePath,
},
promptTemplate: "Follow the paperclip heartbeat.",
},
context: {
issueId: "issue-1",
taskId: "issue-1",
wakeReason: "issue_commented",
wakeCommentId: "comment-2",
paperclipWake: {
reason: "issue_commented",
issue: {
id: "issue-1",
identifier: "PAP-874",
title: "chat-speed issues",
status: "in_progress",
priority: "medium",
},
commentIds: ["comment-1", "comment-2"],
latestCommentId: "comment-2",
comments: [
{
id: "comment-1",
issueId: "issue-1",
body: "First comment",
bodyTruncated: false,
createdAt: "2026-03-28T14:35:00.000Z",
author: { type: "user", id: "user-1" },
},
{
id: "comment-2",
issueId: "issue-1",
body: "Second comment",
bodyTruncated: false,
createdAt: "2026-03-28T14:35:10.000Z",
author: { type: "user", id: "user-1" },
},
],
commentWindow: {
requestedCount: 2,
includedCount: 2,
missingCount: 0,
},
truncated: false,
fallbackFetchNeeded: false,
},
},
authToken: "run-jwt-token",
onLog: async () => {},
});
expect(result.exitCode).toBe(0);
expect(result.errorMessage).toBeNull();
const capture = JSON.parse(await fs.readFile(capturePath, "utf8")) as CapturePayload;
expect(capture.paperclipEnvKeys).toContain("PAPERCLIP_WAKE_PAYLOAD_JSON");
expect(capture.paperclipWakePayloadJson).not.toBeNull();
expect(JSON.parse(capture.paperclipWakePayloadJson ?? "{}")).toMatchObject({
reason: "issue_commented",
latestCommentId: "comment-2",
commentIds: ["comment-1", "comment-2"],
});
expect(capture.prompt).toContain("## Paperclip Wake Payload");
expect(capture.prompt).toContain("Use this wake to continue the task, applying new user direction and preserving its approval gates.");
expect(capture.prompt).toContain("Do not switch to another issue until you have handled this wake.");
expect(capture.prompt).toContain(
"acknowledge the latest comment and explain how it changes your next action.",
);
expect(capture.prompt).toContain("First comment");
expect(capture.prompt).toContain("Second comment");
} finally {
if (previousHome === undefined) delete process.env.HOME;
else process.env.HOME = previousHome;
await fs.rm(root, { recursive: true, force: true });
}
});
it("classifies remote-compaction high-demand failures as retryable transient upstream errors", async () => {
const root = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-codex-execute-transient-"));
const workspace = path.join(root, "workspace");
const commandPath = path.join(root, "codex");
await fs.mkdir(workspace, { recursive: true });
await writeFailingCodexCommand(
commandPath,
"Error running remote compact task: We're currently experiencing high demand, which may cause temporary errors.",
);
const previousHome = process.env.HOME;
process.env.HOME = root;
await seedSharedCodexAuth(root);
try {
const result = await execute({
runId: "run-transient-error",
agent: {
id: "agent-1",
companyId: "company-1",
name: "Codex Coder",
adapterType: "codex_local",
adapterConfig: { engine: "cli" },
},
runtime: {
sessionId: null,
sessionParams: null,
sessionDisplayId: null,
taskKey: null,
},
config: {
engine: "cli",
command: commandPath,
cwd: workspace,
promptTemplate: "Follow the paperclip heartbeat.",
},
context: {},
authToken: "run-jwt-token",
onLog: async () => {},
});
expect(result.exitCode).toBe(1);
expect(result.errorCode).toBe("codex_transient_upstream");
expect(result.errorFamily).toBe("transient_upstream");
expect(result.errorMessage).toContain("high demand");
} finally {
if (previousHome === undefined) delete process.env.HOME;
else process.env.HOME = previousHome;
await fs.rm(root, { recursive: true, force: true });
}
});
it.each([true, false])("retries missing resume only before a session starts (started=%s)", async (started) => {
const root = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-codex-resume-stop-"));
const commandPath = path.join(root, "codex");
const attemptsPath = path.join(root, "attempts");
await seedSharedCodexAuth(root);
await fs.writeFile(commandPath, `#!/usr/bin/env node
const fs = require("node:fs");
fs.appendFileSync(${JSON.stringify(attemptsPath)}, "attempt\\n");
if (process.argv.includes("resume")) {
console.error("state db missing rollout path for thread unrelated-old-thread");
${started ? 'console.log(JSON.stringify({ type: "thread.started", thread_id: "existing-session" }));' : ''}
process.exitCode = 1;
} else {
console.log(JSON.stringify({ type: "thread.started", thread_id: "fresh-session" }));
console.log(JSON.stringify({ type: "turn.completed", usage: { input_tokens: 1, output_tokens: 1 } }));
}
`, "utf8");
await fs.chmod(commandPath, 0o755);
try {
const result = await execute({
runId: `resume-stop-${started}`,
agent: { id: "agent-1", companyId: "company-1", name: "Codex", adapterType: "codex_local", adapterConfig: { engine: "cli" } },
runtime: { sessionId: "existing-session", sessionParams: null, sessionDisplayId: "existing-session", taskKey: null },
config: { engine: "cli", command: commandPath, cwd: root, promptTemplate: "Test resume." },
context: {}, onLog: async () => {},
});
expect((await fs.readFile(attemptsPath, "utf8")).trim().split("\n")).toHaveLength(started ? 1 : 2);
expect(result.sessionId).toBe(started ? "existing-session" : "fresh-session");
} finally {
await fs.rm(root, { recursive: true, force: true });
}
});
it("classifies mid-turn harness crashes as retryable transient upstream errors", async () => {
const root = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-codex-execute-harness-crash-"));
const workspace = path.join(root, "workspace");
const commandPath = path.join(root, "codex");
await fs.mkdir(workspace, { recursive: true });
// Faithful to the observed MCP transport crash: the protocol stream starts,
// then the process dies with only a harness tracing line on stderr — no
// protocol-terminal event (error / turn.failed / turn.completed).
const script = `#!/usr/bin/env node
console.log(JSON.stringify({ type: "thread.started", thread_id: "thread-crash-1" }));
console.log(JSON.stringify({ type: "item.completed", item: { type: "agent_message", text: "Starting the task." } }));
console.error("2026-07-23T22:58:56.007042Z ERROR rmcp::transport::worker: worker quit with fatal: Transport channel closed, when UnexpectedContentType(Some(\\"text/plain\\"))");
process.exit(1);
`;
await fs.writeFile(commandPath, script, "utf8");
await fs.chmod(commandPath, 0o755);
const previousHome = process.env.HOME;
process.env.HOME = root;
await seedSharedCodexAuth(root);
try {
const result = await execute({
runId: "run-harness-crash",
agent: {
id: "agent-1",
companyId: "company-1",
name: "Codex Coder",
adapterType: "codex_local",
adapterConfig: { engine: "cli" },
},
runtime: {
sessionId: null,
sessionParams: null,
sessionDisplayId: null,
taskKey: null,
},
config: {
engine: "cli",
command: commandPath,
cwd: workspace,
promptTemplate: "Follow the paperclip heartbeat.",
},
context: {},
authToken: "run-jwt-token",
onLog: async () => {},
});
expect(result.exitCode).toBe(1);
expect(result.errorCode).toBe("codex_harness_crash");
expect(result.errorFamily).toBe("transient_upstream");
expect(result.errorMessage).toContain("Transport channel closed");
expect(result.sessionId).toBe("thread-crash-1");
expect(result.clearSession).toBe(false);
expect((result.resultJson as Record<string, unknown>).errorFamily).toBe("transient_upstream");
} finally {
if (previousHome === undefined) delete process.env.HOME;
else process.env.HOME = previousHome;
await fs.rm(root, { recursive: true, force: true });
}
});
it("persists retry-not-before metadata for codex provider quota failures", async () => {
const root = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-codex-execute-usage-limit-"));
const workspace = path.join(root, "workspace");
const commandPath = path.join(root, "codex");
await fs.mkdir(workspace, { recursive: true });
await writeFailingCodexCommand(
commandPath,
"You've hit your usage limit for GPT-5.3-Codex-Spark. Switch to another model now, or try again at 11:31 PM.",
);
const previousHome = process.env.HOME;
process.env.HOME = root;
await seedSharedCodexAuth(root);
vi.useFakeTimers();
vi.setSystemTime(new Date(2026, 3, 22, 22, 29, 0));
try {
const result = await execute({
runId: "run-usage-limit",
agent: {
id: "agent-1",
companyId: "company-1",
name: "Codex Coder",
adapterType: "codex_local",
adapterConfig: { engine: "cli" },
},
runtime: {
sessionId: "codex-session-usage-limit",
sessionParams: {
sessionId: "codex-session-usage-limit",
cwd: workspace,
},
sessionDisplayId: "codex-session-usage-limit",
taskKey: null,
},
config: {
engine: "cli",
command: commandPath,
cwd: workspace,
model: "gpt-5.3-codex-spark",
promptTemplate: "Follow the paperclip heartbeat.",
},
context: {},
authToken: "run-jwt-token",
onLog: async () => {},
});
expect(result.exitCode).toBe(1);
expect(result.errorCode).toBe("provider_quota");
expect(result.errorFamily).toBe("provider_quota");
const expectedRetryNotBefore = new Date(2026, 3, 22, 23, 31, 0, 0).toISOString();
expect(result.retryNotBefore).toBe(expectedRetryNotBefore);
expect(result.resultJson?.retryNotBefore).toBe(expectedRetryNotBefore);
expect(result.resultJson?.providerQuotaRetryNotBefore).toBe(expectedRetryNotBefore);
expect(new Date(String(result.resultJson?.transientRetryNotBefore)).getTime()).toBe(
new Date(2026, 3, 22, 23, 31, 0, 0).getTime(),
);
} finally {
vi.useRealTimers();
if (previousHome === undefined) delete process.env.HOME;
else process.env.HOME = previousHome;
await fs.rm(root, { recursive: true, force: true });
}
});
it("classifies Codex refresh-token auth failures without credential telemetry", async () => {
const root = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-codex-execute-refresh-token-"));
const workspace = path.join(root, "workspace");
const commandPath = path.join(root, "codex");
await fs.mkdir(workspace, { recursive: true });
await writeFailingCodexCommand(commandPath, "OAuth failed: refresh_token_reused");
const previousHome = process.env.HOME;
process.env.HOME = root;
await seedSharedCodexAuth(root);
try {
const result = await execute({
runId: "run-refresh-token-reused",
agent: {
id: "agent-1",
companyId: "company-1",
name: "Codex Coder",
adapterType: "codex_local",
adapterConfig: { engine: "cli" },
},
runtime: {
sessionId: null,
sessionParams: null,
sessionDisplayId: null,
taskKey: null,
},
config: {
engine: "cli",
command: commandPath,
cwd: workspace,
promptTemplate: "Follow the paperclip heartbeat.",
},
context: {},
authToken: "run-jwt-token",
onLog: async () => {},
});
expect(result.exitCode).toBe(1);
expect(result.errorCode).toBe("refresh_token_reused");
expect(result.errorFamily).toBe("refresh_token_reused");
expect(result.resultJson?.errorFamily).toBe("refresh_token_reused");
expect(result.resultJson).not.toHaveProperty("codexCredentialTelemetry");
} finally {
if (previousHome === undefined) delete process.env.HOME;
else process.env.HOME = previousHome;
await fs.rm(root, { recursive: true, force: true });
}
});
it("uses safer invocation settings and a fresh-session handoff for codex transient fallback retries", async () => {
const root = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-codex-execute-fallback-"));
const workspace = path.join(root, "workspace");
const commandPath = path.join(root, "codex");
const capturePath = path.join(root, "capture.json");
await fs.mkdir(workspace, { recursive: true });
await writeFakeCodexCommand(commandPath);
const previousHome = process.env.HOME;
process.env.HOME = root;
await seedSharedCodexAuth(root);
let commandNotes: string[] = [];
try {
const result = await execute({
runId: "run-fallback",
agent: {
id: "agent-1",
companyId: "company-1",
name: "Codex Coder",
adapterType: "codex_local",
adapterConfig: { engine: "cli" },
},
runtime: {
sessionId: null,
sessionParams: {
sessionId: "codex-session-stale",
cwd: workspace,
},
sessionDisplayId: "codex-session-stale",
taskKey: null,
},
config: {
engine: "cli",
command: commandPath,
cwd: workspace,
fastMode: true,
model: "gpt-5.4",
env: {
PAPERCLIP_TEST_CAPTURE_PATH: capturePath,
},
promptTemplate: "Follow the paperclip heartbeat.",
},
context: {
codexTransientFallbackMode: "fresh_session_safer_invocation",
paperclipContinuationSummary: {
key: "continuation-summary",
title: "Continuation Summary",
body: "Issue continuation summary for the next fresh session.",
updatedAt: "2026-04-21T01:00:00.000Z",
},
},
authToken: "run-jwt-token",
onLog: async () => {},
onMeta: async (meta) => {
commandNotes = meta.commandNotes ?? [];
},
});
expect(result.exitCode).toBe(0);
expect(result.errorMessage).toBeNull();
const capture = JSON.parse(await fs.readFile(capturePath, "utf8")) as CapturePayload;
expect(capture.argv).toEqual(expect.arrayContaining(["exec", "--json", "-"]));
expect(capture.argv).not.toContain("resume");
expect(capture.argv).not.toContain('service_tier="fast"');
expect(capture.argv).not.toContain("features.fast_mode=true");
expect(capture.prompt).toContain("Paperclip session handoff:");
expect(capture.prompt).toContain("Issue continuation summary for the next fresh session.");
expect(commandNotes).toContain("Codex transient fallback requested safer invocation settings for this retry.");
expect(commandNotes).toContain("Codex transient fallback forced a fresh session with a continuation handoff.");
} finally {
if (previousHome === undefined) delete process.env.HOME;
else process.env.HOME = previousHome;
await fs.rm(root, { recursive: true, force: true });
}
});
it("renders execution-stage wake instructions for reviewer and executor roles", async () => {
const root = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-codex-execute-stage-wake-"));
const workspace = path.join(root, "workspace");
const commandPath = path.join(root, "codex");
const capturePath = path.join(root, "capture.json");
await fs.mkdir(workspace, { recursive: true });
await writeFakeCodexCommand(commandPath);
const previousHome = process.env.HOME;
process.env.HOME = root;
await seedSharedCodexAuth(root);
try {
const result = await execute({
runId: "run-stage-wake",
agent: {
id: "agent-1",
companyId: "company-1",
name: "Codex Coder",
adapterType: "codex_local",
adapterConfig: { engine: "cli" },
},
runtime: {
sessionId: null,
sessionParams: null,
sessionDisplayId: null,
taskKey: null,
},
config: {
engine: "cli",
command: commandPath,
cwd: workspace,
env: {
PAPERCLIP_TEST_CAPTURE_PATH: capturePath,
},
promptTemplate: "Follow the paperclip heartbeat.",
},
context: {
issueId: "issue-1",
taskId: "issue-1",
wakeReason: "execution_review_requested",
paperclipWake: {
reason: "execution_review_requested",
issue: {
id: "issue-1",
identifier: "PAP-1207",
title: "implement the plan of PAP-1200",
status: "in_review",
priority: "medium",
},
executionStage: {
wakeRole: "reviewer",
stageId: "stage-1",
stageType: "review",
currentParticipant: { type: "agent", agentId: "qa-agent" },
returnAssignee: { type: "agent", agentId: "coder-agent" },
lastDecisionOutcome: null,
allowedActions: ["approve", "request_changes"],
},
commentIds: [],
latestCommentId: null,
comments: [],
commentWindow: {
requestedCount: 0,
includedCount: 0,
missingCount: 0,
},
truncated: false,
fallbackFetchNeeded: false,
},
},
authToken: "run-jwt-token",
onLog: async () => {},
});
expect(result.exitCode).toBe(0);
const capture = JSON.parse(await fs.readFile(capturePath, "utf8")) as CapturePayload;
expect(capture.prompt).toContain("execution wake role: reviewer");
expect(capture.prompt).toContain("You are waking as the active reviewer for this issue.");
expect(capture.prompt).toContain("Do not execute the task itself or continue executor work.");
expect(capture.prompt).toContain("allowed actions: approve, request_changes");
const executorCapturePath = path.join(root, "capture-executor.json");
const executorResult = await execute({
runId: "run-stage-wake-executor",
agent: {
id: "agent-1",
companyId: "company-1",
name: "Codex Coder",
adapterType: "codex_local",
adapterConfig: { engine: "cli" },
},
runtime: {
sessionId: null,
sessionParams: null,
sessionDisplayId: null,
taskKey: null,
},
config: {
engine: "cli",
command: commandPath,
cwd: workspace,
env: {
PAPERCLIP_TEST_CAPTURE_PATH: executorCapturePath,
},
promptTemplate: "Follow the paperclip heartbeat.",
},
context: {
issueId: "issue-1",
taskId: "issue-1",
wakeReason: "execution_changes_requested",
paperclipWake: {
reason: "execution_changes_requested",
issue: {
id: "issue-1",
identifier: "PAP-1207",
title: "implement the plan of PAP-1200",
status: "in_progress",
priority: "medium",
},
executionStage: {
wakeRole: "executor",
stageId: "stage-1",
stageType: "review",
currentParticipant: { type: "agent", agentId: "qa-agent" },
returnAssignee: { type: "agent", agentId: "coder-agent" },
lastDecisionOutcome: "changes_requested",
allowedActions: ["address_changes", "resubmit"],
},
commentIds: [],
latestCommentId: null,
comments: [],
commentWindow: {
requestedCount: 0,
includedCount: 0,
missingCount: 0,
},
truncated: false,
fallbackFetchNeeded: false,
},
},
authToken: "run-jwt-token",
onLog: async () => {},
});
expect(executorResult.exitCode).toBe(0);
const executorCapture = JSON.parse(await fs.readFile(executorCapturePath, "utf8")) as CapturePayload;
expect(executorCapture.prompt).toContain("execution wake role: executor");
expect(executorCapture.prompt).toContain("You are waking because changes were requested in the execution workflow.");
expect(executorCapture.prompt).toContain("allowed actions: address_changes, resubmit");
} finally {
if (previousHome === undefined) delete process.env.HOME;
else process.env.HOME = previousHome;
await fs.rm(root, { recursive: true, force: true });
}
});
it("renders an issue-scoped wake prompt even when the wake has no comments yet", async () => {
const root = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-codex-execute-issue-wake-"));
const workspace = path.join(root, "workspace");
const commandPath = path.join(root, "codex");
const capturePath = path.join(root, "capture.json");
await fs.mkdir(workspace, { recursive: true });
await writeFakeCodexCommand(commandPath);
const previousHome = process.env.HOME;
process.env.HOME = root;
await seedSharedCodexAuth(root);
try {
const result = await execute({
runId: "run-issue-wake",
agent: {
id: "agent-1",
companyId: "company-1",
name: "Codex Coder",
adapterType: "codex_local",
adapterConfig: { engine: "cli" },
},
runtime: {
sessionId: null,
sessionParams: null,
sessionDisplayId: null,
taskKey: null,
},
config: {
engine: "cli",
command: commandPath,
cwd: workspace,
env: {
PAPERCLIP_TEST_CAPTURE_PATH: capturePath,
},
promptTemplate: "Follow the paperclip heartbeat.",
},
context: {
issueId: "issue-1",
taskId: "issue-1",
wakeReason: "issue_assigned",
paperclipWake: {
reason: "issue_assigned",
issue: {
id: "issue-1",
identifier: "PAP-1201",
title: "Fix gallery opening for inline images",
status: "in_progress",
priority: "medium",
},
checkedOutByHarness: true,
commentIds: [],
latestCommentId: null,
comments: [],
commentWindow: {
requestedCount: 0,
includedCount: 0,
missingCount: 0,
},
truncated: false,
fallbackFetchNeeded: false,
},
},
authToken: "run-jwt-token",
onLog: async () => {},
});
expect(result.exitCode).toBe(0);
expect(result.errorMessage).toBeNull();
const capture = JSON.parse(await fs.readFile(capturePath, "utf8")) as CapturePayload;
expect(capture.paperclipEnvKeys).toContain("PAPERCLIP_WAKE_PAYLOAD_JSON");
expect(capture.paperclipWakePayloadJson).not.toBeNull();
expect(JSON.parse(capture.paperclipWakePayloadJson ?? "{}")).toMatchObject({
reason: "issue_assigned",
issue: {
identifier: "PAP-1201",
title: "Fix gallery opening for inline images",
status: "in_progress",
priority: "medium",
},
checkedOutByHarness: true,
commentIds: [],
});
expect(capture.prompt).toContain("## Paperclip Wake Payload");
expect(capture.prompt).toContain("Do not switch to another issue until you have handled this wake.");
expect(capture.prompt).toContain("- issue: PAP-1201 Fix gallery opening for inline images");
expect(capture.prompt).not.toContain("- pending comments:");
expect(capture.prompt).not.toContain("acknowledge the latest comment");
expect(capture.prompt).not.toContain("Execution contract:");
expect(capture.prompt).toContain("- issue status: in_progress");
expect(capture.prompt).toContain("- checkout: already claimed by the harness for this run");
expect(capture.prompt).toContain("The harness already checked out this issue for the current run.");
} finally {
if (previousHome === undefined) delete process.env.HOME;
else process.env.HOME = previousHome;
await fs.rm(root, { recursive: true, force: true });
}
});
it.each([{ conversationMode: false, resumedSession: true }, { conversationMode: true, resumedSession: true }, { conversationMode: true, resumedSession: false }])("retains current task policy (conversation=$conversationMode, resumed=$resumedSession)", async ({ conversationMode, resumedSession }) => {
const root = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-codex-execute-resume-wake-"));
const workspace = path.join(root, "workspace");
const commandPath = path.join(root, "codex");
const capturePath = path.join(root, "capture.json");
const instructionsPath = path.join(root, "AGENTS.md");
await fs.mkdir(workspace, { recursive: true });
await fs.writeFile(instructionsPath, "You are managed instructions.\n", "utf8");
await writeFakeCodexCommand(commandPath);
const previousHome = process.env.HOME;
process.env.HOME = root;
await seedSharedCodexAuth(root);
const policy = conversationMode
? buildPaperclipTaskMarkdown({
issue: { id: "issue-1", title: "Chat", workMode: "planning", conversationAgentId: "agent-1" },
interaction: { kind: "request_confirmation", status: "rejected" },
planReview: { status: "rejected", reason: "Revise the final note." },
includeDescription: false,
})
: "Current ordinary task policy";
let invocationPrompt = "";
let invocationNotes: string[] = [];
let promptMetrics: Record<string, number> = {};
try {
const result = await execute({
runId: "run-resume-wake",
agent: {
id: "agent-1",
companyId: "company-1",
name: "Codex Coder",
adapterType: "codex_local",
adapterConfig: { engine: "cli" },
},
runtime: {
sessionId: null,
sessionParams: {
sessionId: resumedSession ? "codex-session-1" : null,
cwd: workspace,
},
sessionDisplayId: null,
taskKey: null,
},
config: {
engine: "cli",
command: commandPath,
cwd: workspace,
instructionsFilePath: instructionsPath,
env: {
PAPERCLIP_TEST_CAPTURE_PATH: capturePath,
},
promptTemplate: conversationMode ? undefined : "Follow the paperclip heartbeat.",
},
context: {
conversationMode,
paperclipTaskMarkdown: `Full description that must not replay\n${policy}`,
paperclipTaskMarkdownCompact: policy,
issueId: "issue-1",
taskId: "issue-1",
wakeReason: "issue_commented",
wakeCommentId: "comment-2",
paperclipWake: {
reason: "issue_commented",
issue: {
id: "issue-1",
identifier: "PAP-874",
title: "chat-speed issues",
status: "in_progress",
priority: "medium",
},
commentIds: ["comment-2"],
latestCommentId: "comment-2",
comments: [
{
id: "comment-2",
issueId: "issue-1",
body: "Second comment",
bodyTruncated: false,
createdAt: "2026-03-28T14:35:10.000Z",
author: { type: "user", id: "user-1" },
},
],
commentWindow: {
requestedCount: 1,
includedCount: 1,
missingCount: 0,
},
truncated: false,
fallbackFetchNeeded: false,
},
},
authToken: "run-jwt-token",
onLog: async () => {},
onMeta: async (meta) => {
invocationPrompt = meta.prompt ?? "";
invocationNotes = meta.commandNotes ?? [];
promptMetrics = meta.promptMetrics ?? {};
},
});
expect(result.exitCode).toBe(0);
expect(result.errorMessage).toBeNull();
const capture = JSON.parse(await fs.readFile(capturePath, "utf8")) as CapturePayload;
if (resumedSession) expect(capture.argv).toEqual(expect.arrayContaining(["resume", "codex-session-1", "-"]));
else expect(capture.argv).not.toContain("resume");
expect(capture.prompt).toContain(resumedSession ? "## Paperclip Resume Delta" : "## Paperclip Wake Payload");
expect(capture.prompt).toContain("Do not switch to another issue until you have handled this wake.");
expect(capture.prompt).toContain("Second comment");
expect(capture.prompt).toContain(policy);
expect(invocationPrompt).toContain(policy);
if (resumedSession) expect(capture.prompt).not.toContain("Full description that must not replay");
else expect(capture.prompt).toContain("Full description that must not replay");
expect(promptMetrics.taskContextChars).toBe(resumedSession ? policy.length : `Full description that must not replay\n${policy}`.length);
if (conversationMode) {
expect(invocationPrompt).toContain(AGENT_CHAT_DIRECTIVE);
expect(invocationPrompt).toContain("baseRevisionId set to that latestRevisionId");
expect(capture.prompt).not.toContain("Execution contract:");
expect(capture.prompt).not.toContain("Use child issues");
} else {
expect(capture.prompt).toContain("Execution contract:");
}
expect(capture.prompt).not.toContain("Follow the paperclip heartbeat.");
if (resumedSession) {
expect(capture.prompt).not.toContain("You are managed instructions.");
expect(invocationPrompt).toContain("## Paperclip Resume Delta");
expect(invocationNotes).toContain("Skipped stdin instruction reinjection because an existing Codex session is being resumed with a wake delta.");
expect(promptMetrics.instructionsChars).toBe(0);
expect(promptMetrics.heartbeatPromptChars).toBe(0);
} else {
expect(capture.prompt).toContain("You are managed instructions.");
expect(promptMetrics.heartbeatPromptChars).toBeGreaterThan(0);
}
} finally {
if (previousHome === undefined) delete process.env.HOME;
else process.env.HOME = previousHome;
await fs.rm(root, { recursive: true, force: true });
}
});
it("uses a worktree-isolated CODEX_HOME and mounts the operational skill by default", async () => {
const root = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-codex-execute-"));
const workspace = path.join(root, "workspace");
const commandPath = path.join(root, "codex");
const capturePath = path.join(root, "capture.json");
const sharedCodexHome = path.join(root, "shared-codex-home");
const paperclipHome = path.join(root, "paperclip-home");
const isolatedCodexHome = path.join(
paperclipHome,
"instances",
"worktree-1",
"companies",
"company-1",
"codex-home",
);
const homeSkill = path.join(isolatedCodexHome, "skills", "paperclip");
await fs.mkdir(workspace, { recursive: true });
await fs.mkdir(sharedCodexHome, { recursive: true });
await fs.writeFile(path.join(sharedCodexHome, "auth.json"), `${fakeCodexAuthJson}\n`, "utf8");
await fs.writeFile(path.join(sharedCodexHome, "config.toml"), 'model = "codex-mini-latest"\n', "utf8");
await writeFakeCodexCommand(commandPath);
const previousHome = process.env.HOME;
const previousPaperclipHome = process.env.PAPERCLIP_HOME;
const previousPaperclipInstanceId = process.env.PAPERCLIP_INSTANCE_ID;
const previousPaperclipInWorktree = process.env.PAPERCLIP_IN_WORKTREE;
const previousCodexHome = process.env.CODEX_HOME;
process.env.HOME = root;
process.env.PAPERCLIP_HOME = paperclipHome;
process.env.PAPERCLIP_INSTANCE_ID = "worktree-1";
process.env.PAPERCLIP_IN_WORKTREE = "true";
process.env.CODEX_HOME = sharedCodexHome;
try {
const logs: LogEntry[] = [];
const result = await execute({
runId: "run-1",
agent: {
id: "agent-1",
companyId: "company-1",
name: "Codex Coder",
adapterType: "codex_local",
adapterConfig: { engine: "cli" },
},
runtime: {
sessionId: null,
sessionParams: null,
sessionDisplayId: null,
taskKey: null,
},
config: {
engine: "cli",
command: commandPath,
cwd: workspace,
env: {
PAPERCLIP_TEST_CAPTURE_PATH: capturePath,
},
promptTemplate: "Follow the paperclip heartbeat.",
},
context: {},
authToken: "run-jwt-token",
onLog: async (stream, chunk) => {
logs.push({ stream, chunk });
},
});
expect(result.exitCode).toBe(0);
expect(result.errorMessage).toBeNull();
const capture = JSON.parse(await fs.readFile(capturePath, "utf8")) as CapturePayload;
expect(capture.codexHome).toBe(isolatedCodexHome);
expect(capture.argv).toEqual(expect.arrayContaining(["exec", "--json", "-"]));
expect(capture.prompt).toContain("Follow the paperclip heartbeat.");
expect(capture.paperclipEnvKeys).toEqual(
expect.arrayContaining([
"PAPERCLIP_AGENT_ID",
"PAPERCLIP_API_KEY",
"PAPERCLIP_API_URL",
"PAPERCLIP_COMPANY_ID",
"PAPERCLIP_RUN_ID",
]),
);
const isolatedAuth = path.join(isolatedCodexHome, "auth.json");
const isolatedConfig = path.join(isolatedCodexHome, "config.toml");
expect((await fs.lstat(isolatedAuth)).isSymbolicLink()).toBe(true);
expect(await fs.realpath(isolatedAuth)).toBe(await fs.realpath(path.join(sharedCodexHome, "auth.json")));
expect((await fs.lstat(isolatedConfig)).isFile()).toBe(true);
expect(await fs.readFile(isolatedConfig, "utf8")).toBe('model = "codex-mini-latest"\n');
expect((await fs.lstat(homeSkill)).isSymbolicLink()).toBe(true);
expect(logs).toContainEqual(
expect.objectContaining({
stream: "stdout",
chunk: expect.stringContaining("Using worktree-isolated Codex home"),
}),
);
expect(logs).toContainEqual(
expect.objectContaining({
stream: "stdout",
chunk: expect.stringContaining('Injected Codex skill "paperclip"'),
}),
);
} finally {
if (previousHome === undefined) delete process.env.HOME;
else process.env.HOME = previousHome;
if (previousPaperclipHome === undefined) delete process.env.PAPERCLIP_HOME;
else process.env.PAPERCLIP_HOME = previousPaperclipHome;
if (previousPaperclipInstanceId === undefined) delete process.env.PAPERCLIP_INSTANCE_ID;
else process.env.PAPERCLIP_INSTANCE_ID = previousPaperclipInstanceId;
if (previousPaperclipInWorktree === undefined) delete process.env.PAPERCLIP_IN_WORKTREE;
else process.env.PAPERCLIP_IN_WORKTREE = previousPaperclipInWorktree;
if (previousCodexHome === undefined) delete process.env.CODEX_HOME;
else process.env.CODEX_HOME = previousCodexHome;
await fs.rm(root, { recursive: true, force: true });
}
});
it("isolates connector skills by agent and revision without changing the selected model identity", async () => {
const root = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-connector-codex-"));
const workspace = path.join(root, "workspace");
const command = path.join(root, "codex");
const capture = path.join(root, "capture.json");
const sourceHome = path.join(root, "selected-account");
const skillSource = path.join(root, "skill-v1");
await fs.mkdir(workspace, { recursive: true });
await fs.mkdir(sourceHome, { recursive: true });
await fs.mkdir(skillSource, { recursive: true });
await fs.writeFile(path.join(sourceHome, "auth.json"), fakeCodexAuthJson);
await fs.writeFile(path.join(skillSource, "SKILL.md"), "# AgentMail\nAssigned inbox one.");
await writeFakeCodexCommand(command);
const keys = ["PAPERCLIP_HOME", "PAPERCLIP_INSTANCE_ID", "CODEX_HOME"] as const;
const previous = Object.fromEntries(keys.map((key) => [key, process.env[key]]));
process.env.PAPERCLIP_HOME = path.join(root, "paperclip");
process.env.PAPERCLIP_INSTANCE_ID = "connectors";
process.env.CODEX_HOME = sourceHome;
const invoke = async (agentId: string, digest: string | null, source = skillSource, connectorSkillInstructions = "") => {
const config = {
engine: "cli", command, cwd: workspace,
env: { CODEX_HOME: sourceHome, PAPERCLIP_TEST_CAPTURE_PATH: capture },
paperclipConnectorSkillDigest: digest,
paperclipSkillSync: { desiredSkills: digest ? ["paperclipai/paperclip/agentmail"] : [] },
paperclipRuntimeSkills: digest ? [{ key: "paperclipai/paperclip/agentmail", runtimeName: "agentmail", source }] : [],
};
const result = await execute({ runId: `run-${agentId}-${digest?.slice(0, 1) ?? "none"}`,
agent: { id: agentId, companyId: "company-1", name: "Email agent", adapterType: "codex_local", adapterConfig: config },
runtime: { sessionId: null, sessionParams: null, sessionDisplayId: null, taskKey: null },
config, context: { paperclipWake: { connectorSkillInstructions } }, authToken: "test-token", onLog: async () => {},
});
expect(result.errorMessage).toBeNull();
expect(result.exitCode).toBe(0);
return JSON.parse(await fs.readFile(capture, "utf8")) as CapturePayload;
};
try {
const first = await invoke("agent-1", "a".repeat(64));
expect(first.codexHome).toContain("connector-runtimes/agent-1/");
expect(await fs.realpath(path.join(first.codexHome!, "auth.json"))).toBe(await fs.realpath(path.join(sourceHome, "auth.json")));
expect(await fs.readFile(path.join(first.codexHome!, "skills/agentmail/SKILL.md"), "utf8")).toContain("inbox one");
await expect(fs.stat(path.join(sourceHome, "skills/agentmail"))).rejects.toMatchObject({ code: "ENOENT" });
const other = await invoke("agent-2", "a".repeat(64));
expect(other.codexHome).not.toBe(first.codexHome);
const nextSource = path.join(root, "skill-v2");
await fs.mkdir(nextSource);
await fs.writeFile(path.join(nextSource, "SKILL.md"), "# AgentMail\nAssigned inbox two.");
const next = await invoke("agent-1", "b".repeat(64), nextSource);
expect(next.codexHome).not.toBe(first.codexHome);
expect(await fs.readFile(path.join(next.codexHome!, "skills/agentmail/SKILL.md"), "utf8")).toContain("inbox two");
const inline = await invoke("agent-1", null, skillSource, "# AgentMail\nAssigned inbox inline@example.test");
expect(inline.prompt).toContain("Assigned inbox inline@example.test");
await expect(fs.stat(path.join(sourceHome, "skills/agentmail"))).rejects.toMatchObject({ code: "ENOENT" });
const removed = await invoke("agent-1", null);
expect(removed.prompt).not.toContain("inline@example.test");
expect(removed.codexHome).toBe(sourceHome);
await expect(fs.stat(path.join(removed.codexHome!, "skills/agentmail"))).rejects.toMatchObject({ code: "ENOENT" });
} finally {
for (const key of keys) { if (previous[key] === undefined) delete process.env[key]; else process.env[key] = previous[key]; }
await fs.rm(root, { recursive: true, force: true });
}
});
it("respects an explicit CODEX_HOME config override even in worktree mode", async () => {
const root = await fs.mkdtemp(path.join(os.tmpdir(), "paperclip-codex-execute-explicit-"));
const workspace = path.join(root, "workspace");
const commandPath = path.join(root, "codex");
const capturePath = path.join(root, "capture.json");
const sharedCodexHome = path.join(root, "shared-codex-home");
const explicitCodexHome = path.join(root, "explicit-codex-home");
const paperclipHome = path.join(root, "paperclip-home");
await fs.mkdir(workspace, { recursive: true });
await fs.mkdir(sharedCodexHome, { recursive: true });
await fs.writeFile(path.join(sharedCodexHome, "auth.json"), `${fakeCodexAuthJson}\n`, "utf8");
await writeFakeCodexCommand(commandPath);
const previousHome = process.env.HOME;
const previousPaperclipHome = process.env.PAPERCLIP_HOME;
const previousPaperclipInstanceId = process.env.PAPERCLIP_INSTANCE_ID;
const previousPaperclipInWorktree = process.env.PAPERCLIP_IN_WORKTREE;
const previousCodexHome = process.env.CODEX_HOME;
process.env.HOME = root;
process.env.PAPERCLIP_HOME = paperclipHome;
process.env.PAPERCLIP_INSTANCE_ID = "worktree-1";
process.env.PAPERCLIP_IN_WORKTREE = "true";
process.env.CODEX_HOME = sharedCodexHome;
try {
const result = await execute({
runId: "run-2",
agent: {
id: "agent-1",
companyId: "company-1",
name: "Codex Coder",
adapterType: "codex_local",
adapterConfig: { engine: "cli" },
},
runtime: {
sessionId: null,
sessionParams: null,
sessionDisplayId: null,
taskKey: null,
},
config: {
engine: "cli",
command: commandPath,
cwd: workspace,
env: {
PAPERCLIP_TEST_CAPTURE_PATH: capturePath,
CODEX_HOME: explicitCodexHome,
},
promptTemplate: "Follow the paperclip heartbeat.",
paperclipSkillSync: {
desiredSkills: ["paperclip"],
},
},
context: {},
authToken: "run-jwt-token",
onLog: async () => {},
});
expect(result.exitCode).toBe(0);
expect(result.errorMessage).toBeNull();
const capture = JSON.parse(await fs.readFile(capturePath, "utf8")) as CapturePayload;
expect(capture.codexHome).toBe(explicitCodexHome);
expect((await fs.lstat(path.join(explicitCodexHome, "skills", "paperclip"))).isSymbolicLink()).toBe(true);
await expect(fs.lstat(path.join(paperclipHome, "instances", "worktree-1", "codex-home"))).rejects.toThrow();
} finally {
if (previousHome === undefined) delete process.env.HOME;
else process.env.HOME = previousHome;
if (previousPaperclipHome === undefined) delete process.env.PAPERCLIP_HOME;
else process.env.PAPERCLIP_HOME = previousPaperclipHome;
if (previousPaperclipInstanceId === undefined) delete process.env.PAPERCLIP_INSTANCE_ID;
else process.env.PAPERCLIP_INSTANCE_ID = previousPaperclipInstanceId;
if (previousPaperclipInWorktree === undefined) delete process.env.PAPERCLIP_IN_WORKTREE;
else process.env.PAPERCLIP_IN_WORKTREE = previousPaperclipInWorktree;
if (previousCodexHome === undefined) delete process.env.CODEX_HOME;
else process.env.CODEX_HOME = previousCodexHome;
await fs.rm(root, { recursive: true, force: true });
}
});
});