181 lines
5.1 KiB
TypeScript
181 lines
5.1 KiB
TypeScript
import { join } from "node:path";
|
|
import { tmpdir } from "node:os";
|
|
import { mkdtemp, rm } from "node:fs/promises";
|
|
import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
|
|
|
|
vi.mock("electron", () => ({
|
|
safeStorage: {
|
|
isEncryptionAvailable: () => false,
|
|
getSelectedStorageBackend: () => "basic_text",
|
|
encryptString: (value: string) => Buffer.from(value, "utf8"),
|
|
decryptString: (value: Buffer) => value.toString("utf8"),
|
|
},
|
|
}));
|
|
|
|
import { AppDatabase } from "../../src/main/database";
|
|
import {
|
|
handleBeforeToolCall,
|
|
type ActiveRunContext,
|
|
} from "../../src/main/agent-runtime/hook-handlers";
|
|
|
|
function seedRun(db: AppDatabase, executionMode: "ai_ceo" | "ai_cofounder" = "ai_cofounder") {
|
|
const companyId = db.core.company.saveCompany({
|
|
name: "Gate Co",
|
|
description: "Cofounder runtime gate test",
|
|
status: "active",
|
|
executionMode,
|
|
});
|
|
db.core.connectors.saveConnector({
|
|
id: "openai-codex",
|
|
model: "gpt-5",
|
|
envBindingText: "",
|
|
notes: "fixture",
|
|
});
|
|
const agentId = db.core.agents.saveAgent({
|
|
companyId,
|
|
name: "Researcher",
|
|
role: "Research Lead",
|
|
title: "Research Lead",
|
|
department: "product",
|
|
status: "active",
|
|
reportsTo: null,
|
|
connectorId: "openai-codex",
|
|
workspaceId: null,
|
|
model: null,
|
|
capabilities: "Research",
|
|
budgetMonthlyUsd: 100,
|
|
});
|
|
const taskId = db.core.tasks.saveTask({
|
|
companyId,
|
|
title: "Draft market plan",
|
|
description: "Prepare the market plan after approval.",
|
|
assigneeAgentId: agentId,
|
|
priority: "high",
|
|
status: "in_progress",
|
|
requiresFilesystem: false,
|
|
});
|
|
const runContext: ActiveRunContext = {
|
|
runId: "run-1",
|
|
agentId,
|
|
companyId,
|
|
taskId,
|
|
connectorId: "openai-codex",
|
|
modelId: "gpt-5",
|
|
taskSessionId: "session-1",
|
|
workspaceId: null,
|
|
workspacePath: null,
|
|
workspaceMode: "none",
|
|
workspaceBaseline: null,
|
|
state: "agent_active",
|
|
};
|
|
return { companyId, agentId, taskId, runContext };
|
|
}
|
|
|
|
describe("cofounder approval runtime gate", () => {
|
|
let tempDir: string;
|
|
let db: AppDatabase;
|
|
|
|
beforeEach(async () => {
|
|
tempDir = await mkdtemp(join(tmpdir(), "cofounder-gate-"));
|
|
db = new AppDatabase(tempDir);
|
|
db.init();
|
|
});
|
|
|
|
afterEach(async () => {
|
|
db.close();
|
|
await rm(tempDir, { recursive: true, force: true });
|
|
});
|
|
|
|
it("blocks destructive tools while a cofounder plan review is pending", () => {
|
|
const { companyId, agentId, taskId, runContext } = seedRun(db, "ai_cofounder");
|
|
const approvalId = db.core.approvals.requestApproval({
|
|
companyId,
|
|
relatedTaskId: taskId,
|
|
requestedByAgentId: agentId,
|
|
type: "plan_review",
|
|
payloadSummary: "Review plan",
|
|
impactSummary: "Human approval is required before execution.",
|
|
payloadJson: "{}",
|
|
});
|
|
|
|
const blocking = db.core.approvals.getBlockingApproval(companyId, agentId, taskId);
|
|
expect(blocking?.id).toBe(approvalId);
|
|
|
|
const result = handleBeforeToolCall(
|
|
{
|
|
db,
|
|
logger: { info: () => {}, warn: () => {}, error: () => {} },
|
|
},
|
|
runContext,
|
|
{
|
|
toolName: "bash",
|
|
toolCallId: "call-1",
|
|
args: { command: "touch blocked.txt" },
|
|
},
|
|
);
|
|
|
|
expect(result.block).toBe(true);
|
|
expect(result.reason).toContain("plan_review");
|
|
});
|
|
|
|
it("lets CEO-mode execution proceed past plan review policy", () => {
|
|
const { companyId, agentId, taskId, runContext } = seedRun(db, "ai_ceo");
|
|
db.core.approvals.requestApproval({
|
|
companyId,
|
|
relatedTaskId: taskId,
|
|
requestedByAgentId: agentId,
|
|
type: "plan_review",
|
|
payloadSummary: "Review plan",
|
|
impactSummary: "Human review requested for planning.",
|
|
payloadJson: "{}",
|
|
});
|
|
|
|
expect(db.core.approvals.getBlockingApproval(companyId, agentId, taskId)).toBeNull();
|
|
const result = handleBeforeToolCall(
|
|
{
|
|
db,
|
|
logger: { info: () => {}, warn: () => {}, error: () => {} },
|
|
},
|
|
runContext,
|
|
{
|
|
toolName: "bash",
|
|
toolCallId: "call-1",
|
|
args: { command: "touch allowed.txt" },
|
|
},
|
|
);
|
|
|
|
expect(result.block).toBe(false);
|
|
});
|
|
|
|
it("blocks operational approvals in CEO mode", () => {
|
|
const { companyId, agentId, taskId, runContext } = seedRun(db, "ai_ceo");
|
|
const approvalId = db.core.approvals.requestApproval({
|
|
companyId,
|
|
relatedTaskId: taskId,
|
|
requestedByAgentId: agentId,
|
|
type: "dangerous_command",
|
|
payloadSummary: "Approve destructive command",
|
|
impactSummary: "Host-level mutation requires explicit approval.",
|
|
payloadJson: "{}",
|
|
});
|
|
|
|
const blocking = db.core.approvals.getBlockingApproval(companyId, agentId, taskId);
|
|
expect(blocking?.id).toBe(approvalId);
|
|
const result = handleBeforeToolCall(
|
|
{
|
|
db,
|
|
logger: { info: () => {}, warn: () => {}, error: () => {} },
|
|
},
|
|
runContext,
|
|
{
|
|
toolName: "bash",
|
|
toolCallId: "call-1",
|
|
args: { command: "rm -rf dist" },
|
|
},
|
|
);
|
|
|
|
expect(result.block).toBe(true);
|
|
expect(result.reason).toContain("dangerous_command");
|
|
});
|
|
});
|