mirror of
https://github.com/paperclipai/paperclip.git
synced 2026-11-04 02:32:07 +01:00
## Thinking Path > - Paperclip manages AI agents and their paid work. > - Budgets must prevent new work and stop reported overruns. > - Cost settlement also ran serial lifecycle reconciliation for every eligible company agent. > - That made request latency grow with unrelated idle agents and delayed results on stop delivery. > - This change separates atomic admission and accounting from durable asynchronous enforcement. > - Scoped live usage detects overruns without writing provisional ledger charges. > - The result is constant query scope for ordinary settlement and recoverable stop delivery. ## Linked Issues or Issue Description **What happened?** Settlement and budget checks entered company-wide lifecycle reconciliation. A company with 1,000 unrelated idle agents performed 1,001 accounting transactions for one ordinary cost evaluation. User requests could also wait for provider termination. Running usage checkpoints did not enforce reported monetary exposure. **Expected behavior** Keep the final reservation gate authoritative. Commit charges, scope blocks, and enforcement intent before returning. Deliver stop requests after commit. Read only applicable policies and execution scopes for ordinary charges. Stop long-running work when supported reported or priced usage establishes an overrun. **Steps to reproduce** 1. Create a disposable company with one company budget and idle agents. 2. Evaluate a charge below the budget with 0, 17, and 1,000 unrelated idle agents. 3. Compare query counts, company accounting transactions, and duration. 4. Use the controlled partial-usage and delayed-stop fixtures to check enforcement and request responsiveness. **Paperclip version or commit** The measured baseline is `6c9b87d0922f1ebfc43e8b0c36889ac5ba7b9837` from `master`. Related public work: Refs #16976, #2006, #3539, #13763, #14296, #16853, #13075. These cover the affected service charge, cancellation, and accounting paths. This change removes their shared reconciliation overhead. ## What Changed - Batch applicable policy exposure under the existing company accounting lock. Preserve exact fractional money and zero-reservation dispatch fences. - Commit a durable final budget dispatch claim before provider I/O. Retain a separate final issue/run ownership transaction through synchronous handoff, without holding the company accounting lock or awaiting provider work. Preserve native recovery lease checks and duplicate handoff fencing. Browser handoff rechecks current policy; it reuses a consumed rate slot only for the identical policy generation. - Separate ledger spend, live observed spend, and effective held capacity. Use the maximum of reservation and live observation for each unfinished operation. - Deliver budget cancellation through the existing coordinator. Revalidate generations and persist queued/running stop fences before external termination. Bound stop concurrency to four. - Apply agent holds only to the affected agent. Project budgets leave other projects running. Project real company transitions in bounded lifecycle batches. - Replace periodic broad reconciliation with durable pending work, startup repair, receipt replay, and concrete retry/window deadlines. Retain drain and idle-safety checks. - Enforce accepted cumulative usage checkpoints without provisional charges. Final accounting atomically replaces live exposure and settles reservations. - Add operator-gated budget spans, regression coverage, map updates, and a real-PostgreSQL comparison harness. No schema or first-party telemetry change. ## Verification - Full workspace `pnpm -r typecheck` passes on the final ownership fix, rebased onto `bbcf5dfe98`. Production `pnpm build` passes on final head `4c5f4de8b1`. Map, module-boundary, token, and Storybook catalog checks pass. - Focused real-PostgreSQL coverage includes the 292-case rebased accounting/dispatch run, 63 browser cases, 52 run-control/cancellation cases, 42 policy-generation cases, 77 lifecycle/commentary cases, and the exact Muse confirmation regression. Native executor unit coverage passes all 768 cases. Final ownership coverage passes: 74 adapter/use-case cases, nine public regression cases, and the additional deleted-project race. These exercise competing ownership, budget changes and charges between phases, zero/nonzero duplicate claims, rollback, recovery contention, read-only capacity rechecks, and controlled HTTP startup after commit. Another 74 pure/startup/tracing cases pass. The complete CI matrix and both 5/5 reviews passed on `1ac9b58`; its external security scan identified a stale-owner handoff race. The final fix restores synchronous issue/run ownership fencing after the accounting transaction commits. Final head `8e84f55ce8` has fresh Greptile 5/5 with zero new comments, Paperclip review 5/5, and a passing security scan. The complete latest-head [CI matrix](https://github.com/paperclipai/paperclip/actions/runs/39535189808) passes, including budget, browser, Runner, build, and canary jobs. The unchanged Runner close fixture passed on its sole same-head job retry. - Broad local `pnpm test:run` invocations were stopped before required source updates (rebase and review fixes). They are not claimed as full passes. Final full-suite coverage comes from the complete latest-head CI matrix; local focused coverage and the incomplete broad invocations are reported separately. - Reproduce performance with `pnpm --filter @paperclipai/server exec tsx ../scripts/benchmark-budget-enforcement.ts --baseline-ref=6c9b87d0922f1ebfc43e8b0c36889ac5ba7b9837`; repeat with `--reverse`. The harness uses disposable real PostgreSQL, 20 warm samples per variant/count, and records raw timings, SQL/lock counts, agent reads, unchanged agent state, and query plans. Final production-source measurement revision: `eff713a4f72c7f4b7a4002a701cb481d4bfccd95`; the extended harness is committed in `4c5f4de8b1`. | Unrelated idle agents | Baseline median / p95 | New median / p95 | Baseline SQL / locks | New SQL / locks | New agent reads | | --- | ---: | ---: | ---: | ---: | ---: | | 0 | 1.80 / 2.33 ms | 2.99 / 3.88 ms | 5 / 1 | 7 / 1 | 0 | | 17 | 15.53 / 19.55 ms | 2.24 / 2.60 ms | 90 / 18 | 7 / 1 | 0 | | 1,000 | 978.35 / 1,790.43 ms | 2.16 / 2.52 ms | 5,005 / 1,001 | 7 / 1 | 0 | - These measurements use local PostgreSQL 18.1 and Node 26.4.0 arm64. They isolate the cost-evaluation core and complete admission path, without hosted-region or provider latency. Prior-revision reversed-order controls also preserved the constant query counts; the final measurements above use the final production source. - Complete reserve → final ownership gate → immediate offline provider result is measured separately, with one responsible agent and one active operation at each unrelated-agent count. Admission p95 baseline/new: 0 agents **11.54 / 29.77 ms**; 17 agents **9.31 / 27.62 ms**; 1,000 agents **24.51 / 74.05 ms**. Baseline admission was already scoped; the stronger final fence adds bounded overhead. New counts stay constant at 53 Drizzle statements, three transactions, two company-accounting locks, and five agent-reading statements bound to the responsible agent. Zero unrelated-agent reads, exactly one provider callback per sample, and no leftover runs or holds. All admission scenarios stay below 250 ms. No external provider calls are included. - The partial-usage/delayed-termination fixture passed 20/20: receipt persistence p95 13 ms, receipt return to observed durable Stop fence p95 28 ms, complete receipt-start-to-fence p95 50 ms, unrelated database request p95 1 ms while termination was blocked. After releasing the controlled termination fixture, acknowledgement p95 was 14 ms. These are separately measured phases; fixture acknowledgement is not live-provider termination latency. - Models and provider deadlines are unchanged. No deployment, live inference, or staging acceptance is claimed. A quiet 30-minute follow-up waits for the merged revision on staging, then repeats classifier and controlled budget-stop acceptance in staging, recording hosted bookkeeping, queue delay, and actual provider acknowledgement separately. - Development failures retained: finalizer notification regressions (fixed; 27/27 affected cases passed), invalid browser admission fixture assumptions (fixed; browser coverage now passes 63/63), and first-head lifecycle/dispatch CI regressions (reproduced and repaired). A prior-head Cases E2E test hit its unchanged announcement-dismiss handler race while opening a filter; the same journey passed on `1ac9b58`. The external security review then found a stale-owner race between accounting commit and provider entry; final verification covers a competing run in that gap while retaining provider I/O outside the accounting lock. Early PostgreSQL startup attempts hit local semaphore pressure; serial reruns succeeded. The final-head Runner lane had one failure among 1,823 tests: the unchanged `timeout-attachment` fixture failed at its first process close after 6.28 seconds. The sole same-head job retry passed all 1,819 active tests (four skipped), including that fixture in 1.85 seconds. Three isolated local probes could not run correctly because the Rust fake-provider fixture binary was absent; cleanup masked the original startup failure. CI builds that fixture through its normal lane. These probes are inconclusive and are not counted as passes. ## Risks - This changes accounting, lifecycle, and cancellation ordering. Durable generations, source/sequence checks, ownership leases, and company locking remain required for correctness. - Provider stop acknowledgement can be slow or fail. Stop intent remains pending until existing termination proof succeeds; cancelled row status alone does not acknowledge delivery. - Mid-run limits depend on provider usage reports or supported price calculations. Providers that report only at completion cannot provide exact mid-call detection. - Signals are process-local wake hints for the supported single-active-writer model. Durable state and startup recovery remain authoritative. - There are no migrations, new queue tables, or cached balances. Repeat the classifier and controlled budget-stop acceptance after normal rollout. ## Model Used OpenAI Codex, based on GPT-6, with reasoning, tool use, code execution, and parallel coding agents. The host does not expose the exact backend model revision or context-window size, so those values are not asserted. ## 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 - [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>
1653 lines
64 KiB
TypeScript
1653 lines
64 KiB
TypeScript
import { randomUUID } from "node:crypto";
|
|
import { eq } from "drizzle-orm";
|
|
import { afterAll, afterEach, beforeAll, describe, expect, it } from "vitest";
|
|
import {
|
|
activityLog,
|
|
agents,
|
|
companies,
|
|
companySecrets,
|
|
createDb,
|
|
heartbeatRuns,
|
|
issues,
|
|
principalPermissionGrants,
|
|
projects,
|
|
toolAccessAuditEvents,
|
|
toolActionRequests,
|
|
toolApplications,
|
|
toolCatalogEntries,
|
|
toolCallEvents,
|
|
toolConnections,
|
|
toolInvocations,
|
|
toolMcpGateways,
|
|
toolPolicies,
|
|
toolProfileBindings,
|
|
toolProfileEntries,
|
|
toolProfiles,
|
|
toolRateLimitCounters,
|
|
} from "@paperclipai/db";
|
|
import {
|
|
getEmbeddedPostgresTestSupport,
|
|
startEmbeddedPostgresTestDatabase,
|
|
} from "./helpers/embedded-postgres.js";
|
|
import { toolAccessPolicyService } from "../services/tool-access-policy.js";
|
|
import { toolAccessService } from "../services/tool-access.js";
|
|
import { createToolGatewayService, ToolGatewayHttpError } from "../services/tool-gateway.js";
|
|
|
|
const embeddedPostgresSupport = await getEmbeddedPostgresTestSupport();
|
|
const describeEmbeddedPostgres = embeddedPostgresSupport.supported ? describe : describe.skip;
|
|
|
|
async function createCompany(db: ReturnType<typeof createDb>) {
|
|
return db.insert(companies).values({
|
|
name: `Tool Access ${randomUUID()}`,
|
|
issuePrefix: `TA${randomUUID().slice(0, 6).toUpperCase()}`,
|
|
}).returning().then((rows) => rows[0]!);
|
|
}
|
|
|
|
async function createAgent(
|
|
db: ReturnType<typeof createDb>,
|
|
companyId: string,
|
|
permissions: Record<string, unknown> = {},
|
|
) {
|
|
return db.insert(agents).values({
|
|
companyId,
|
|
name: `Agent ${randomUUID()}`,
|
|
role: "engineer",
|
|
adapterType: "process",
|
|
adapterConfig: {},
|
|
runtimeConfig: {},
|
|
permissions,
|
|
}).returning().then((rows) => rows[0]!);
|
|
}
|
|
|
|
async function createRun(
|
|
db: ReturnType<typeof createDb>,
|
|
companyId: string,
|
|
agentId: string,
|
|
contextSnapshot: Record<string, unknown> = {},
|
|
) {
|
|
return db.insert(heartbeatRuns).values({
|
|
companyId,
|
|
agentId,
|
|
invocationSource: "assignment",
|
|
status: "running",
|
|
contextSnapshot,
|
|
}).returning().then((rows) => rows[0]!);
|
|
}
|
|
|
|
async function createIssue(db: ReturnType<typeof createDb>, companyId: string, title = "Tool issue") {
|
|
return db.insert(issues).values({
|
|
companyId,
|
|
title: `${title} ${randomUUID()}`,
|
|
status: "in_progress",
|
|
}).returning().then((rows) => rows[0]!);
|
|
}
|
|
|
|
async function createTool(db: ReturnType<typeof createDb>, companyId: string) {
|
|
const application = await db.insert(toolApplications).values({
|
|
companyId,
|
|
applicationKey: `fixture-${randomUUID()}`,
|
|
name: `Fixture ${randomUUID()}`,
|
|
type: "mcp_http",
|
|
status: "active",
|
|
}).returning().then((rows) => rows[0]!);
|
|
const connection = await db.insert(toolConnections).values({
|
|
companyId,
|
|
applicationId: application.id,
|
|
name: `Connection ${randomUUID()}`,
|
|
uid: `test/${randomUUID()}`,
|
|
transport: "mcp_remote",
|
|
status: "active",
|
|
enabled: true,
|
|
config: { url: "https://example.invalid/mcp" },
|
|
}).returning().then((rows) => rows[0]!);
|
|
const catalogEntry = await db.insert(toolCatalogEntries).values({
|
|
companyId,
|
|
applicationId: application.id,
|
|
connectionId: connection.id,
|
|
name: "send_email",
|
|
toolName: "send_email",
|
|
riskLevel: "write",
|
|
versionHash: randomUUID(),
|
|
schemaHash: randomUUID(),
|
|
}).returning().then((rows) => rows[0]!);
|
|
return { application, connection, catalogEntry };
|
|
}
|
|
|
|
async function createApprovedToolAction(input: {
|
|
db: ReturnType<typeof createDb>;
|
|
companyId: string;
|
|
agentId: string;
|
|
connectionId: string;
|
|
catalogEntryId: string;
|
|
issueId?: string | null;
|
|
argumentsValue: Record<string, unknown>;
|
|
status?: "approved" | "executed";
|
|
}) {
|
|
const svc = toolAccessPolicyService(input.db);
|
|
const decisionInput = {
|
|
companyId: input.companyId,
|
|
actor: { actorType: "agent" as const, actorId: input.agentId, agentId: input.agentId },
|
|
runContext: { issueId: input.issueId ?? null },
|
|
request: {
|
|
connectionId: input.connectionId,
|
|
catalogEntryId: input.catalogEntryId,
|
|
toolName: "send_email",
|
|
arguments: input.argumentsValue,
|
|
},
|
|
};
|
|
const decision = await svc.decide(decisionInput);
|
|
const recorded = await svc.recordInvocation(decisionInput, decision);
|
|
if (!recorded.actionRequest) throw new Error("Expected approval-required action request");
|
|
const [updated] = await input.db
|
|
.update(toolActionRequests)
|
|
.set({
|
|
status: input.status ?? "approved",
|
|
resolvedAt: new Date(),
|
|
updatedAt: new Date(),
|
|
})
|
|
.where(eq(toolActionRequests.id, recorded.actionRequest.id))
|
|
.returning();
|
|
return { decisionInput, invocation: recorded.invocation, actionRequest: updated };
|
|
}
|
|
|
|
type IdRef = { id: string };
|
|
|
|
function sendEmailGrantRow(company: IdRef, agent: IdRef): typeof principalPermissionGrants.$inferInsert {
|
|
return {
|
|
companyId: company.id,
|
|
principalType: "agent",
|
|
principalId: agent.id,
|
|
permissionKey: "tools:use",
|
|
scope: { toolName: "send_email" },
|
|
};
|
|
}
|
|
|
|
function agentProfileBindingRow(company: IdRef, profile: IdRef, agent: IdRef): typeof toolProfileBindings.$inferInsert {
|
|
return {
|
|
companyId: company.id,
|
|
profileId: profile.id,
|
|
targetType: "agent",
|
|
targetId: agent.id,
|
|
};
|
|
}
|
|
|
|
function sendEmailDecisionInput(company: IdRef, agent: IdRef, connection: IdRef, catalogEntry: IdRef): Parameters<ReturnType<typeof toolAccessPolicyService>["decide"]>[0] {
|
|
return {
|
|
companyId: company.id,
|
|
actor: { actorType: "agent", actorId: agent.id, agentId: agent.id },
|
|
request: { connectionId: connection.id, catalogEntryId: catalogEntry.id, toolName: "send_email" },
|
|
};
|
|
}
|
|
|
|
function reviewSendPolicyRow(company: IdRef, name: string): typeof toolPolicies.$inferInsert {
|
|
return {
|
|
companyId: company.id,
|
|
name: name,
|
|
policyType: "require_approval",
|
|
selectors: { toolName: "send_email" },
|
|
};
|
|
}
|
|
|
|
function executedActionInput(db: ReturnType<typeof createDb>, company: IdRef, agent: IdRef, connection: IdRef, catalogEntry: IdRef, args: Record<string, unknown>): Parameters<typeof createApprovedToolAction>[0] {
|
|
return {
|
|
db,
|
|
companyId: company.id,
|
|
agentId: agent.id,
|
|
connectionId: connection.id,
|
|
catalogEntryId: catalogEntry.id,
|
|
argumentsValue: args,
|
|
status: "executed",
|
|
};
|
|
}
|
|
|
|
describeEmbeddedPostgres("tool access policy service", () => {
|
|
let db!: ReturnType<typeof createDb>;
|
|
let tempDb: Awaited<ReturnType<typeof startEmbeddedPostgresTestDatabase>> | null = null;
|
|
|
|
beforeAll(async () => {
|
|
tempDb = await startEmbeddedPostgresTestDatabase("paperclip-tool-access-policy-");
|
|
db = createDb(tempDb.connectionString);
|
|
}, 20_000);
|
|
|
|
afterEach(async () => {
|
|
await db.delete(toolRateLimitCounters);
|
|
await db.delete(toolActionRequests);
|
|
await db.delete(toolInvocations);
|
|
await db.delete(toolCallEvents);
|
|
await db.delete(toolAccessAuditEvents);
|
|
await db.delete(toolPolicies);
|
|
await db.delete(toolProfileEntries);
|
|
await db.delete(toolProfileBindings);
|
|
await db.delete(toolMcpGateways);
|
|
await db.delete(toolProfiles);
|
|
await db.delete(toolCatalogEntries);
|
|
await db.delete(toolConnections);
|
|
await db.delete(toolApplications);
|
|
await db.delete(companySecrets);
|
|
await db.delete(principalPermissionGrants);
|
|
await db.delete(issues);
|
|
await db.delete(projects);
|
|
await db.delete(activityLog);
|
|
await db.delete(heartbeatRuns);
|
|
await db.delete(agents);
|
|
await db.delete(companies);
|
|
});
|
|
|
|
afterAll(async () => {
|
|
await tempDb?.cleanup();
|
|
});
|
|
|
|
it("denies direct execution without an effective profile or grant", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
|
|
const result = await toolAccessPolicyService(db).decide(sendEmailDecisionInput(company, agent, connection, catalogEntry));
|
|
|
|
expect(result).toMatchObject({
|
|
allowed: false,
|
|
decision: "deny",
|
|
reasonCode: "deny_default",
|
|
});
|
|
});
|
|
|
|
it("allows calls through an effective agent profile", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
const profile = await db.insert(toolProfiles).values({
|
|
companyId: company.id,
|
|
profileKey: `profile-${randomUUID()}`,
|
|
name: "Write tools",
|
|
defaultAction: "deny",
|
|
}).returning().then((rows) => rows[0]!);
|
|
await db.insert(toolProfileBindings).values(agentProfileBindingRow(company, profile, agent));
|
|
await db.insert(toolProfileEntries).values({
|
|
companyId: company.id,
|
|
profileId: profile.id,
|
|
selectorType: "tool_name",
|
|
effect: "include",
|
|
toolName: "send_email",
|
|
});
|
|
|
|
const result = await toolAccessPolicyService(db).decide(sendEmailDecisionInput(company, agent, connection, catalogEntry));
|
|
|
|
expect(result).toMatchObject({
|
|
allowed: true,
|
|
decision: "allow",
|
|
reasonCode: "allow_profile",
|
|
effectiveProfileIds: [profile.id],
|
|
});
|
|
});
|
|
|
|
it("uses issue-scoped profiles ahead of company defaults", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
const issue = await createIssue(db, company.id, "Scoped deny");
|
|
|
|
const [companyProfile, issueProfile] = await db.insert(toolProfiles).values([
|
|
{
|
|
companyId: company.id,
|
|
profileKey: `company-allow-${randomUUID()}`,
|
|
name: "Company allow",
|
|
defaultAction: "deny",
|
|
},
|
|
{
|
|
companyId: company.id,
|
|
profileKey: `issue-deny-${randomUUID()}`,
|
|
name: "Issue deny",
|
|
defaultAction: "deny",
|
|
},
|
|
]).returning();
|
|
await db.insert(toolProfileBindings).values([
|
|
{
|
|
companyId: company.id,
|
|
profileId: companyProfile!.id,
|
|
targetType: "company",
|
|
targetId: company.id,
|
|
},
|
|
{
|
|
companyId: company.id,
|
|
profileId: issueProfile!.id,
|
|
targetType: "issue",
|
|
targetId: issue.id,
|
|
},
|
|
]);
|
|
await db.insert(toolProfileEntries).values({
|
|
companyId: company.id,
|
|
profileId: companyProfile!.id,
|
|
selectorType: "tool_name",
|
|
effect: "include",
|
|
toolName: "send_email",
|
|
});
|
|
|
|
const result = await toolAccessPolicyService(db).decide({
|
|
companyId: company.id,
|
|
actor: { actorType: "agent", actorId: agent.id, agentId: agent.id },
|
|
runContext: { issueId: issue.id },
|
|
request: { connectionId: connection.id, catalogEntryId: catalogEntry.id, toolName: "send_email" },
|
|
});
|
|
|
|
expect(result).toMatchObject({
|
|
allowed: false,
|
|
decision: "deny",
|
|
reasonCode: "deny_default",
|
|
effectiveProfileIds: [issueProfile!.id],
|
|
});
|
|
});
|
|
|
|
it("uses agent-scoped profiles ahead of project defaults", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
const project = await db.insert(projects).values({
|
|
companyId: company.id,
|
|
name: `Project ${randomUUID()}`,
|
|
}).returning().then((rows) => rows[0]!);
|
|
|
|
const [projectProfile, agentProfile] = await db.insert(toolProfiles).values([
|
|
{
|
|
companyId: company.id,
|
|
profileKey: `project-allow-${randomUUID()}`,
|
|
name: "Project allow",
|
|
defaultAction: "deny",
|
|
},
|
|
{
|
|
companyId: company.id,
|
|
profileKey: `agent-deny-${randomUUID()}`,
|
|
name: "Agent deny",
|
|
defaultAction: "deny",
|
|
},
|
|
]).returning();
|
|
await db.insert(toolProfileBindings).values([
|
|
{
|
|
companyId: company.id,
|
|
profileId: projectProfile!.id,
|
|
targetType: "project",
|
|
targetId: project.id,
|
|
},
|
|
{
|
|
companyId: company.id,
|
|
profileId: agentProfile!.id,
|
|
targetType: "agent",
|
|
targetId: agent.id,
|
|
},
|
|
]);
|
|
await db.insert(toolProfileEntries).values({
|
|
companyId: company.id,
|
|
profileId: projectProfile!.id,
|
|
selectorType: "tool_name",
|
|
effect: "include",
|
|
toolName: "send_email",
|
|
});
|
|
|
|
const result = await toolAccessPolicyService(db).decide({
|
|
companyId: company.id,
|
|
actor: { actorType: "agent", actorId: agent.id, agentId: agent.id },
|
|
runContext: { projectId: project.id },
|
|
request: { connectionId: connection.id, catalogEntryId: catalogEntry.id, toolName: "send_email" },
|
|
});
|
|
|
|
expect(result).toMatchObject({
|
|
allowed: false,
|
|
decision: "deny",
|
|
reasonCode: "deny_default",
|
|
effectiveProfileIds: [agentProfile!.id],
|
|
});
|
|
});
|
|
|
|
it("uses issue-scoped allows ahead of broader company denies", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
const issue = await createIssue(db, company.id, "Scoped allow");
|
|
|
|
const [companyProfile, issueProfile] = await db.insert(toolProfiles).values([
|
|
{
|
|
companyId: company.id,
|
|
profileKey: `company-deny-${randomUUID()}`,
|
|
name: "Company deny",
|
|
defaultAction: "deny",
|
|
},
|
|
{
|
|
companyId: company.id,
|
|
profileKey: `issue-allow-${randomUUID()}`,
|
|
name: "Issue allow",
|
|
defaultAction: "deny",
|
|
},
|
|
]).returning();
|
|
await db.insert(toolProfileBindings).values([
|
|
{
|
|
companyId: company.id,
|
|
profileId: companyProfile!.id,
|
|
targetType: "company",
|
|
targetId: company.id,
|
|
},
|
|
{
|
|
companyId: company.id,
|
|
profileId: issueProfile!.id,
|
|
targetType: "issue",
|
|
targetId: issue.id,
|
|
},
|
|
]);
|
|
await db.insert(toolProfileEntries).values({
|
|
companyId: company.id,
|
|
profileId: issueProfile!.id,
|
|
selectorType: "tool_name",
|
|
effect: "include",
|
|
toolName: "send_email",
|
|
});
|
|
|
|
const result = await toolAccessPolicyService(db).decide({
|
|
companyId: company.id,
|
|
actor: { actorType: "agent", actorId: agent.id, agentId: agent.id },
|
|
runContext: { issueId: issue.id },
|
|
request: { connectionId: connection.id, catalogEntryId: catalogEntry.id, toolName: "send_email" },
|
|
});
|
|
|
|
expect(result).toMatchObject({
|
|
allowed: true,
|
|
decision: "allow",
|
|
reasonCode: "allow_profile",
|
|
effectiveProfileIds: [issueProfile!.id],
|
|
});
|
|
});
|
|
|
|
it("denies calls through draft and archived profiles", async () => {
|
|
const company = await createCompany(db);
|
|
const draftAgent = await createAgent(db, company.id);
|
|
const archivedAgent = await createAgent(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
const [draftProfile, archivedProfile] = await db.insert(toolProfiles).values([
|
|
{
|
|
companyId: company.id,
|
|
profileKey: `draft-profile-${randomUUID()}`,
|
|
name: "Draft write tools",
|
|
status: "draft",
|
|
defaultAction: "allow",
|
|
},
|
|
{
|
|
companyId: company.id,
|
|
profileKey: `archived-profile-${randomUUID()}`,
|
|
name: "Archived write tools",
|
|
status: "archived",
|
|
defaultAction: "allow",
|
|
},
|
|
]).returning();
|
|
await db.insert(toolProfileBindings).values([
|
|
{
|
|
companyId: company.id,
|
|
profileId: draftProfile!.id,
|
|
targetType: "agent",
|
|
targetId: draftAgent.id,
|
|
},
|
|
{
|
|
companyId: company.id,
|
|
profileId: archivedProfile!.id,
|
|
targetType: "agent",
|
|
targetId: archivedAgent.id,
|
|
},
|
|
]);
|
|
await db.insert(toolProfileEntries).values([
|
|
{
|
|
companyId: company.id,
|
|
profileId: draftProfile!.id,
|
|
selectorType: "tool_name",
|
|
effect: "include",
|
|
toolName: "send_email",
|
|
},
|
|
{
|
|
companyId: company.id,
|
|
profileId: archivedProfile!.id,
|
|
selectorType: "tool_name",
|
|
effect: "include",
|
|
toolName: "send_email",
|
|
},
|
|
]);
|
|
|
|
for (const agent of [draftAgent, archivedAgent]) {
|
|
const result = await toolAccessPolicyService(db).decide(sendEmailDecisionInput(company, agent, connection, catalogEntry));
|
|
expect(result).toMatchObject({
|
|
allowed: false,
|
|
decision: "deny",
|
|
reasonCode: "deny_default",
|
|
effectiveProfileIds: [],
|
|
});
|
|
}
|
|
});
|
|
|
|
it("denies calls through disabled applications before explicit grants and allows them after reactivation", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { application, connection, catalogEntry } = await createTool(db, company.id);
|
|
await db.insert(principalPermissionGrants).values(sendEmailGrantRow(company, agent));
|
|
const input = {
|
|
companyId: company.id,
|
|
actor: { actorType: "agent" as const, actorId: agent.id, agentId: agent.id },
|
|
request: { connectionId: connection.id, catalogEntryId: catalogEntry.id, toolName: "send_email" },
|
|
};
|
|
|
|
await expect(toolAccessPolicyService(db).decide(input)).resolves.toMatchObject({
|
|
allowed: true,
|
|
decision: "allow",
|
|
reasonCode: "allow_explicit_grant",
|
|
});
|
|
|
|
await db
|
|
.update(toolApplications)
|
|
.set({ status: "disabled", updatedAt: new Date() })
|
|
.where(eq(toolApplications.id, application.id));
|
|
await expect(toolAccessPolicyService(db).decide(input)).resolves.toMatchObject({
|
|
allowed: false,
|
|
decision: "deny",
|
|
reasonCode: "deny_disabled_application",
|
|
explanation: "Application is disabled.",
|
|
});
|
|
|
|
await db
|
|
.update(toolApplications)
|
|
.set({ status: "active", updatedAt: new Date() })
|
|
.where(eq(toolApplications.id, application.id));
|
|
await expect(toolAccessPolicyService(db).decide(input)).resolves.toMatchObject({
|
|
allowed: true,
|
|
decision: "allow",
|
|
reasonCode: "allow_explicit_grant",
|
|
});
|
|
});
|
|
|
|
it("manages generic tool policies without exposing trust rules", async () => {
|
|
const company = await createCompany(db);
|
|
const otherCompany = await createCompany(db);
|
|
const svc = toolAccessPolicyService(db);
|
|
const [trustRule] = await db.insert(toolPolicies).values({
|
|
companyId: company.id,
|
|
name: "Promoted trust rule",
|
|
policyType: "trust_rule",
|
|
selectors: { toolName: "send_email" },
|
|
config: { trustRule: { hitCount: 0 } },
|
|
}).returning();
|
|
|
|
const created = await svc.createPolicy(company.id, {
|
|
name: "Block destructive senders",
|
|
description: "Block a dangerous tool family.",
|
|
policyType: "block",
|
|
priority: 10,
|
|
enabled: true,
|
|
selectors: { toolNames: ["send_email", "delete_email"] },
|
|
conditions: null,
|
|
config: null,
|
|
}, { userId: "board-user" });
|
|
|
|
expect(created).toMatchObject({
|
|
companyId: company.id,
|
|
name: "Block destructive senders",
|
|
policyType: "block",
|
|
priority: 10,
|
|
createdByUserId: "board-user",
|
|
});
|
|
|
|
await expect(svc.createPolicy(company.id, {
|
|
name: "Generic trust rule",
|
|
description: null,
|
|
policyType: "trust_rule",
|
|
priority: 100,
|
|
enabled: true,
|
|
selectors: {},
|
|
conditions: null,
|
|
config: null,
|
|
})).rejects.toMatchObject({ status: 422 });
|
|
|
|
await expect(svc.updatePolicy({
|
|
companyId: otherCompany.id,
|
|
policyId: created.id,
|
|
body: { enabled: false },
|
|
})).rejects.toMatchObject({ status: 404 });
|
|
|
|
const listed = await svc.listPolicies(company.id);
|
|
expect(listed.map((policy) => policy.id)).toEqual([created.id]);
|
|
|
|
const updated = await svc.updatePolicy({
|
|
companyId: company.id,
|
|
policyId: created.id,
|
|
body: {
|
|
name: "Require review for destructive senders",
|
|
policyType: "require_approval",
|
|
enabled: false,
|
|
selectors: { toolName: "delete_email" },
|
|
},
|
|
});
|
|
expect(updated).toMatchObject({
|
|
id: created.id,
|
|
name: "Require review for destructive senders",
|
|
policyType: "require_approval",
|
|
enabled: false,
|
|
selectors: { toolName: "delete_email" },
|
|
});
|
|
|
|
const deleted = await svc.deletePolicy({ companyId: company.id, policyId: created.id });
|
|
expect(deleted.id).toBe(created.id);
|
|
await expect(svc.deletePolicy({ companyId: company.id, policyId: trustRule.id }))
|
|
.rejects.toMatchObject({ status: 404 });
|
|
expect(await svc.listPolicies(company.id)).toEqual([]);
|
|
const remainingTrustRules = await svc.listTrustRules(company.id);
|
|
expect(remainingTrustRules.map((policy) => policy.id)).toEqual([trustRule.id]);
|
|
});
|
|
|
|
it("treats glob-looking action-name selectors as exact names", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
const profile = await db.insert(toolProfiles).values({
|
|
companyId: company.id,
|
|
profileKey: `profile-${randomUUID()}`,
|
|
name: "Write tools",
|
|
defaultAction: "allow",
|
|
}).returning().then((rows) => rows[0]!);
|
|
await db.insert(toolProfileBindings).values(agentProfileBindingRow(company, profile, agent));
|
|
const wildcardPolicy = await db.insert(toolPolicies).values({
|
|
companyId: company.id,
|
|
name: "Review wildcard-looking sends",
|
|
policyType: "require_approval",
|
|
priority: 10,
|
|
selectors: { toolName: "*send*" },
|
|
}).returning().then((rows) => rows[0]!);
|
|
|
|
const input = {
|
|
companyId: company.id,
|
|
actor: { actorType: "agent" as const, actorId: agent.id, agentId: agent.id },
|
|
request: { connectionId: connection.id, catalogEntryId: catalogEntry.id, toolName: "send_email" },
|
|
};
|
|
|
|
await expect(toolAccessPolicyService(db).decide(input)).resolves.toMatchObject({
|
|
allowed: true,
|
|
decision: "allow",
|
|
reasonCode: "allow_profile",
|
|
effectiveProfileIds: [profile.id],
|
|
matchedPolicyIds: [],
|
|
});
|
|
|
|
const exactPolicy = await db.insert(toolPolicies).values({
|
|
companyId: company.id,
|
|
name: "Review exact sends",
|
|
policyType: "require_approval",
|
|
priority: 5,
|
|
selectors: { toolName: "send_email" },
|
|
}).returning().then((rows) => rows[0]!);
|
|
|
|
await expect(toolAccessPolicyService(db).decide(input)).resolves.toMatchObject({
|
|
allowed: false,
|
|
decision: "require_approval",
|
|
reasonCode: "requires_approval_policy",
|
|
effectiveProfileIds: [profile.id],
|
|
matchedPolicyIds: [exactPolicy.id],
|
|
});
|
|
expect(wildcardPolicy.selectors).toEqual({ toolName: "*send*" });
|
|
});
|
|
|
|
it("matches tool name selectors against connected MCP upstream tool names", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
const policy = await db.insert(toolPolicies).values({
|
|
companyId: company.id,
|
|
name: "Review upstream todo writes",
|
|
policyType: "require_approval",
|
|
priority: 10,
|
|
selectors: { toolNames: ["todo.add"] },
|
|
}).returning().then((rows) => rows[0]!);
|
|
|
|
await expect(toolAccessPolicyService(db).decide({
|
|
companyId: company.id,
|
|
actor: { actorType: "agent", actorId: agent.id, agentId: agent.id },
|
|
request: {
|
|
connectionId: connection.id,
|
|
catalogEntryId: catalogEntry.id,
|
|
toolName: "mcp.smoke-fixture:todo-add",
|
|
upstreamToolName: "todo.add",
|
|
},
|
|
})).resolves.toMatchObject({
|
|
allowed: false,
|
|
decision: "require_approval",
|
|
reasonCode: "requires_approval_policy",
|
|
matchedPolicyIds: [policy.id],
|
|
});
|
|
});
|
|
|
|
it("rejects agent-supplied run context that belongs to another agent", async () => {
|
|
const company = await createCompany(db);
|
|
const actorAgent = await createAgent(db, company.id);
|
|
const otherAgent = await createAgent(db, company.id);
|
|
const { connection } = await createTool(db, company.id);
|
|
const run = await db.insert(heartbeatRuns).values({
|
|
companyId: company.id,
|
|
agentId: otherAgent.id,
|
|
invocationSource: "assignment",
|
|
status: "running",
|
|
}).returning().then((rows) => rows[0]!);
|
|
|
|
const result = await toolAccessPolicyService(db).decide({
|
|
companyId: company.id,
|
|
actor: { actorType: "agent", actorId: actorAgent.id, agentId: actorAgent.id },
|
|
runContext: { heartbeatRunId: run.id },
|
|
request: { connectionId: connection.id, toolName: "send_email" },
|
|
});
|
|
|
|
expect(result).toMatchObject({
|
|
allowed: false,
|
|
reasonCode: "deny_run_context_mismatch",
|
|
});
|
|
});
|
|
|
|
it("rejects agent-supplied issue context that differs from the stored heartbeat context", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const canonicalIssue = await createIssue(db, company.id, "Canonical");
|
|
const escalatedIssue = await createIssue(db, company.id, "Escalated");
|
|
const { connection } = await createTool(db, company.id);
|
|
const run = await createRun(db, company.id, agent.id, { issueId: canonicalIssue.id });
|
|
|
|
const result = await toolAccessPolicyService(db).decide({
|
|
companyId: company.id,
|
|
actor: { actorType: "agent", actorId: agent.id, agentId: agent.id },
|
|
runContext: { heartbeatRunId: run.id, issueId: escalatedIssue.id },
|
|
request: { connectionId: connection.id, toolName: "send_email" },
|
|
});
|
|
|
|
expect(result).toMatchObject({
|
|
allowed: false,
|
|
reasonCode: "deny_run_context_mismatch",
|
|
});
|
|
});
|
|
|
|
it("audits denied calls without storing secret argument values", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection } = await createTool(db, company.id);
|
|
const input = {
|
|
companyId: company.id,
|
|
actor: { actorType: "agent" as const, actorId: agent.id, agentId: agent.id },
|
|
request: {
|
|
connectionId: connection.id,
|
|
toolName: "send_email",
|
|
arguments: { to: "ops@example.com", apiKey: "sk-test-secret-value-123456" },
|
|
},
|
|
};
|
|
|
|
const decision = await toolAccessPolicyService(db).decide(input);
|
|
await toolAccessPolicyService(db).writeAudit(input, decision);
|
|
const [legacyAudit] = await db.select().from(toolAccessAuditEvents);
|
|
const [callEvent] = await db.select().from(toolCallEvents);
|
|
const serialized = JSON.stringify({ legacy: legacyAudit.details, dedicated: callEvent });
|
|
|
|
expect(decision.reasonCode).toBe("deny_default");
|
|
expect(callEvent).toMatchObject({
|
|
eventType: "policy_decision",
|
|
outcome: "denied",
|
|
reasonCode: "deny_default",
|
|
decision: "deny",
|
|
matchedPolicyIds: [],
|
|
requestHash: expect.any(String),
|
|
});
|
|
expect(serialized).not.toContain("sk-test-secret-value");
|
|
expect(serialized).toContain("[REDACTED]");
|
|
});
|
|
|
|
it("records approval-required invocations and action requests with matched policy IDs", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection } = await createTool(db, company.id);
|
|
const policy = await db.insert(toolPolicies).values({
|
|
companyId: company.id,
|
|
name: "Review writes",
|
|
policyType: "require_approval",
|
|
selectors: { toolName: "send_email" },
|
|
description: "Writes require board review.",
|
|
}).returning().then((rows) => rows[0]!);
|
|
const input = {
|
|
companyId: company.id,
|
|
actor: { actorType: "agent" as const, actorId: agent.id, agentId: agent.id },
|
|
request: {
|
|
connectionId: connection.id,
|
|
toolName: "send_email",
|
|
arguments: { to: "ops@example.com", body: "ship it" },
|
|
sideEffecting: true,
|
|
},
|
|
};
|
|
|
|
const decision = await toolAccessPolicyService(db).decide(input);
|
|
const recorded = await toolAccessPolicyService(db).recordInvocation(input, decision);
|
|
await toolAccessPolicyService(db).writeAudit(input, decision);
|
|
const [callEvent] = await db.select().from(toolCallEvents);
|
|
|
|
expect(decision).toMatchObject({
|
|
allowed: false,
|
|
decision: "require_approval",
|
|
reasonCode: "requires_approval_policy",
|
|
matchedPolicyIds: [policy.id],
|
|
});
|
|
expect(recorded.actionRequest).toMatchObject({
|
|
invocationId: recorded.invocation.id,
|
|
requestedByAgentId: agent.id,
|
|
status: "pending",
|
|
});
|
|
expect(recorded.invocation).toMatchObject({
|
|
approvalState: "pending",
|
|
status: "awaiting_approval",
|
|
matchedPolicyIds: [policy.id],
|
|
});
|
|
expect(callEvent).toMatchObject({
|
|
eventType: "policy_decision",
|
|
outcome: "pending",
|
|
decision: "require_approval",
|
|
matchedPolicyIds: [policy.id],
|
|
requestSummary: expect.objectContaining({ summary: expect.any(String) }),
|
|
});
|
|
});
|
|
|
|
it("replays side-effecting calls with the same idempotency key instead of creating a new invocation", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection } = await createTool(db, company.id);
|
|
await db.insert(principalPermissionGrants).values(sendEmailGrantRow(company, agent));
|
|
const input = {
|
|
companyId: company.id,
|
|
actor: { actorType: "agent" as const, actorId: agent.id, agentId: agent.id },
|
|
request: {
|
|
connectionId: connection.id,
|
|
toolName: "send_email",
|
|
arguments: { to: "ops@example.com" },
|
|
sideEffecting: true,
|
|
idempotencyKey: "send-email-1",
|
|
},
|
|
};
|
|
|
|
const decision = await toolAccessPolicyService(db).decide(input);
|
|
const first = await toolAccessPolicyService(db).recordInvocation(input, decision);
|
|
const replay = await toolAccessPolicyService(db).recordInvocation(input, decision);
|
|
|
|
expect(first.replayed).toBe(false);
|
|
expect(replay.replayed).toBe(true);
|
|
expect(replay.invocation.id).toBe(first.invocation.id);
|
|
});
|
|
|
|
it("derives a canonical idempotency key for side-effecting calls without caller-supplied keys", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection } = await createTool(db, company.id);
|
|
await db.insert(principalPermissionGrants).values(sendEmailGrantRow(company, agent));
|
|
const input = {
|
|
companyId: company.id,
|
|
actor: { actorType: "agent" as const, actorId: agent.id, agentId: agent.id },
|
|
request: {
|
|
connectionId: connection.id,
|
|
toolName: "send_email",
|
|
arguments: { to: "ops@example.com", body: "only once" },
|
|
sideEffecting: true,
|
|
},
|
|
};
|
|
|
|
const decision = await toolAccessPolicyService(db).decide(input);
|
|
const first = await toolAccessPolicyService(db).recordInvocation(input, decision);
|
|
const replay = await toolAccessPolicyService(db).recordInvocation(input, decision);
|
|
|
|
expect(first.replayed).toBe(false);
|
|
expect(first.invocation.idempotencyKey).toMatch(/^side_effect:/);
|
|
expect(replay.replayed).toBe(true);
|
|
expect(replay.invocation.id).toBe(first.invocation.id);
|
|
});
|
|
|
|
it("enforces rate-limit policies before explicit grants", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection } = await createTool(db, company.id);
|
|
await db.insert(principalPermissionGrants).values(sendEmailGrantRow(company, agent));
|
|
await db.insert(toolPolicies).values({
|
|
companyId: company.id,
|
|
name: "One send per minute",
|
|
policyType: "rate_limit",
|
|
selectors: { toolName: "send_email" },
|
|
config: { limit: 1, windowSeconds: 60, keyBy: ["agent", "tool"] },
|
|
});
|
|
const input = {
|
|
companyId: company.id,
|
|
actor: { actorType: "agent" as const, actorId: agent.id, agentId: agent.id },
|
|
request: { connectionId: connection.id, toolName: "send_email" },
|
|
consumeRateLimit: true,
|
|
};
|
|
|
|
const first = await toolAccessPolicyService(db).decide(input);
|
|
const second = await toolAccessPolicyService(db).decide(input);
|
|
|
|
expect(first.allowed).toBe(true);
|
|
expect(second).toMatchObject({
|
|
allowed: false,
|
|
decision: "rate_limited",
|
|
reasonCode: "rate_limited",
|
|
});
|
|
});
|
|
|
|
it("atomically consumes the final rate-limit slot", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection } = await createTool(db, company.id);
|
|
await db.insert(principalPermissionGrants).values(sendEmailGrantRow(company, agent));
|
|
await db.insert(toolPolicies).values({
|
|
companyId: company.id,
|
|
name: "One concurrent send per minute",
|
|
policyType: "rate_limit",
|
|
selectors: { toolName: "send_email" },
|
|
config: { limit: 1, windowSeconds: 60, keyBy: ["agent", "tool"] },
|
|
});
|
|
const input = {
|
|
companyId: company.id,
|
|
actor: { actorType: "agent" as const, actorId: agent.id, agentId: agent.id },
|
|
request: { connectionId: connection.id, toolName: "send_email" },
|
|
consumeRateLimit: true,
|
|
};
|
|
|
|
const decisions = await Promise.all([
|
|
toolAccessPolicyService(db).decide(input),
|
|
toolAccessPolicyService(db).decide(input),
|
|
]);
|
|
|
|
expect(decisions.filter((decision) => decision.allowed)).toHaveLength(1);
|
|
expect(decisions.filter((decision) => decision.reasonCode === "rate_limited")).toHaveLength(1);
|
|
});
|
|
|
|
async function admittedRateLimitedInvocation(limit = 1) {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const issue = await createIssue(db, company.id);
|
|
const run = await createRun(db, company.id, agent.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
const [profile] = await db.insert(toolProfiles).values({ companyId: company.id,
|
|
profileKey: "admitted-gateway", name: "Admitted gateway", defaultAction: "deny" }).returning();
|
|
const [gateway] = await db.insert(toolMcpGateways).values({ companyId: company.id,
|
|
name: "Admitted gateway", slug: "admitted-gateway", profileId: profile.id }).returning();
|
|
await db.insert(principalPermissionGrants).values(sendEmailGrantRow(company, agent));
|
|
const [ratePolicy] = await db.insert(toolPolicies).values({ companyId: company.id, name: "One admitted send",
|
|
policyType: "rate_limit", selectors: { toolName: "send_email" }, priority: 1,
|
|
config: { limit, windowSeconds: 3600, keyBy: ["agent", "tool"] } }).returning();
|
|
const input = { ...sendEmailDecisionInput(company, agent, connection, catalogEntry),
|
|
runContext: { heartbeatRunId: run.id, issueId: issue.id, gatewayId: gateway.id }, consumeRateLimit: true,
|
|
request: { connectionId: connection.id, catalogEntryId: catalogEntry.id, toolName: "send_email", arguments: { to: "recipient@example.test" } } };
|
|
const svc = toolAccessPolicyService(db);
|
|
const initial = await svc.decide(input);
|
|
expect(initial.allowed).toBe(true);
|
|
const { invocation } = await svc.recordInvocation(input, initial);
|
|
expect(invocation.gatewayId).toBe(gateway.id);
|
|
await db.update(toolInvocations).set({ status: "executing" }).where(eq(toolInvocations.id, invocation.id));
|
|
return { company, agent, input: { ...input, consumeRateLimit: false }, invocation, ratePolicy, svc };
|
|
}
|
|
|
|
it.each(["allow", "approved"] as const)("retains the consumed final rate-limit slot for the same executing %s invocation", async authorization => {
|
|
const { input, invocation, ratePolicy, svc } = await admittedRateLimitedInvocation();
|
|
if (authorization === "approved") await db.update(toolInvocations).set({ policyDecision: "require_approval", approvalState: "approved" })
|
|
.where(eq(toolInvocations.id, invocation.id));
|
|
expect(await svc.decide(input)).toMatchObject({ decision: "rate_limited" });
|
|
expect(invocation.policyExplanation?.consumedRateAdmissions).toEqual([
|
|
{ policyId: ratePolicy.id, versionHash: expect.stringMatching(/^[a-f0-9]{64}$/) },
|
|
]);
|
|
const counters = await db.select().from(toolRateLimitCounters);
|
|
expect(await svc.decide(input, { admittedInvocationId: invocation.id })).toMatchObject({ allowed: true });
|
|
expect(await svc.decide(input, { admittedInvocationId: invocation.id })).toMatchObject({ allowed: true });
|
|
expect(await db.select().from(toolRateLimitCounters)).toEqual(counters);
|
|
expect((await db.select().from(toolRateLimitCounters))[0].remaining).toBe(0);
|
|
// Neither a new invocation nor an accidental consuming recheck can borrow
|
|
// another invocation's already-spent slot.
|
|
expect(await svc.decide(input, { admittedInvocationId: randomUUID() })).toMatchObject({ decision: "rate_limited" });
|
|
expect(await svc.decide({ ...input, consumeRateLimit: true }, { admittedInvocationId: invocation.id }))
|
|
.toMatchObject({ decision: "rate_limited" });
|
|
});
|
|
|
|
it("rechecks a newly added rate policy rather than borrowing an older policy's consumed slot", async () => {
|
|
const { company, input, invocation, svc } = await admittedRateLimitedInvocation();
|
|
await svc.createPolicy(company.id, { name: "New send limit", policyType: "rate_limit", priority: 0,
|
|
selectors: { toolName: "send_email" }, config: { limit: 1, windowSeconds: 3600, keyBy: ["agent", "tool"] } });
|
|
// Consume the new rule through normal admission; the exhausted original
|
|
// rule then denies this separate request before any provider dispatch.
|
|
expect(await svc.decide({ ...input, consumeRateLimit: true })).toMatchObject({ decision: "rate_limited" });
|
|
const counters = await db.select().from(toolRateLimitCounters);
|
|
expect(counters).toHaveLength(2);
|
|
expect(await svc.decide(input, { admittedInvocationId: invocation.id })).toMatchObject({ decision: "rate_limited" });
|
|
expect(await db.select().from(toolRateLimitCounters)).toEqual(counters);
|
|
});
|
|
|
|
it("rechecks an edited rate-policy generation without consuming another slot", async () => {
|
|
const { company, input, invocation, ratePolicy, svc } = await admittedRateLimitedInvocation(2);
|
|
expect(await svc.decide({ ...input, consumeRateLimit: true })).toMatchObject({ allowed: true });
|
|
await svc.updatePolicy({ companyId: company.id, policyId: ratePolicy.id,
|
|
body: { config: { limit: 1, windowSeconds: 3600, keyBy: ["agent", "tool"] } } });
|
|
const counters = await db.select().from(toolRateLimitCounters);
|
|
expect(await svc.decide(input, { admittedInvocationId: invocation.id })).toMatchObject({ decision: "rate_limited" });
|
|
expect(await db.select().from(toolRateLimitCounters)).toEqual(counters);
|
|
});
|
|
|
|
it("rechecks rate limits for historical executing invocations without durable consumption proof", async () => {
|
|
const { input, invocation, svc } = await admittedRateLimitedInvocation();
|
|
await db.update(toolInvocations).set({ policyExplanation: null }).where(eq(toolInvocations.id, invocation.id));
|
|
expect(await svc.decide(input, { admittedInvocationId: invocation.id })).toMatchObject({ decision: "rate_limited" });
|
|
});
|
|
|
|
it("reevaluates current blocks, grants, and approval policies for admitted work", async () => {
|
|
const { company, input, invocation, svc } = await admittedRateLimitedInvocation();
|
|
const options = { admittedInvocationId: invocation.id };
|
|
const [block] = await db.insert(toolPolicies).values({ companyId: company.id, name: "Block admitted sends",
|
|
policyType: "block", priority: 0, selectors: { toolName: "send_email" } }).returning();
|
|
expect(await svc.decide(input, options)).toMatchObject({ decision: "deny", reasonCode: "deny_policy_block" });
|
|
await db.delete(toolPolicies).where(eq(toolPolicies.id, block.id));
|
|
await db.delete(principalPermissionGrants).where(eq(principalPermissionGrants.companyId, company.id));
|
|
expect(await svc.decide(input, options)).toMatchObject({ decision: "deny", reasonCode: "deny_default" });
|
|
await db.insert(toolPolicies).values({ companyId: company.id, name: "Require new review",
|
|
policyType: "require_approval", priority: 2, selectors: { toolName: "send_email" } });
|
|
expect(await svc.decide(input, options)).toMatchObject({ decision: "require_approval" });
|
|
});
|
|
|
|
it("does not retain rate admission for a mismatched or no-longer-authorized invocation", async () => {
|
|
const { input, invocation, svc } = await admittedRateLimitedInvocation();
|
|
const options = { admittedInvocationId: invocation.id };
|
|
const mutations: Partial<typeof toolInvocations.$inferInsert>[] = [
|
|
{ actorType: "system" }, { actorId: randomUUID() }, { agentId: null }, { runId: null }, { issueId: null }, { gatewayId: null },
|
|
{ applicationId: null }, { connectionId: null }, { catalogEntryId: null }, { toolName: "different_tool" },
|
|
{ providerType: null }, { applicationKey: null }, { upstreamToolName: null }, { argumentsHash: "different" },
|
|
{ status: "authorized" }, { policyDecision: "deny" }, { policyDecision: "require_approval", approvalState: "pending" },
|
|
];
|
|
for (const mutation of mutations) {
|
|
await db.update(toolInvocations).set(mutation).where(eq(toolInvocations.id, invocation.id));
|
|
expect(await svc.decide(input, options), JSON.stringify(mutation)).toMatchObject({ decision: "rate_limited" });
|
|
await db.update(toolInvocations).set({ ...invocation, status: "executing" }).where(eq(toolInvocations.id, invocation.id));
|
|
}
|
|
expect(await svc.decide({ ...input, request: { ...input.request, arguments: { to: "another@example.test" } } }, options))
|
|
.toMatchObject({ decision: "rate_limited" });
|
|
});
|
|
|
|
it("rejects unsupported policy semantics at create and update time", async () => {
|
|
const company = await createCompany(db);
|
|
const svc = toolAccessPolicyService(db);
|
|
const policy = await svc.createPolicy(company.id, {
|
|
name: "Require review",
|
|
policyType: "require_approval",
|
|
selectors: { toolName: "send_email" },
|
|
});
|
|
|
|
await expect(svc.createPolicy(company.id, {
|
|
name: "Dead redact policy",
|
|
policyType: "redact" as never,
|
|
selectors: { toolName: "send_email" },
|
|
config: { redact: { fields: ["to", "body"] } },
|
|
})).rejects.toThrow("Tool policy type 'redact' is not supported at runtime");
|
|
await expect(svc.createPolicy(company.id, {
|
|
name: "Dead custom check",
|
|
policyType: "validate" as never,
|
|
selectors: { toolName: "send_email" },
|
|
config: { schema: { required: ["body"] } },
|
|
})).rejects.toThrow("Tool policy type 'validate' is not supported at runtime");
|
|
await expect(svc.createPolicy(company.id, {
|
|
name: "Malformed conditional policy",
|
|
policyType: "allow",
|
|
selectors: { toolName: "send_email" },
|
|
conditions: { args: { body: "safe" } } as never,
|
|
})).rejects.toThrow("Tool policy conditions include unsupported runtime semantics");
|
|
await expect(svc.createPolicy(company.id, {
|
|
name: "Ignored approval config",
|
|
policyType: "require_approval",
|
|
selectors: { toolName: "send_email" },
|
|
config: { validate: { required: ["body"] } },
|
|
})).rejects.toThrow("Tool policy type 'require_approval' does not support config");
|
|
await expect(svc.createPolicy(company.id, {
|
|
name: "Invalid rate limit",
|
|
policyType: "rate_limit",
|
|
selectors: { toolName: "send_email" },
|
|
config: { rateLimit: { limit: 0, windowSeconds: 60 } },
|
|
})).rejects.toThrow("Rate-limit policy config requires positive numeric limit and windowSeconds");
|
|
await expect(svc.updatePolicy({
|
|
companyId: company.id,
|
|
policyId: policy.id,
|
|
body: { conditions: { args: { body: "safe" } } as never },
|
|
})).rejects.toThrow("Tool policy conditions include unsupported runtime semantics");
|
|
});
|
|
|
|
it("rejects fieldMatches patterns with nested quantifiers", async () => {
|
|
const company = await createCompany(db);
|
|
const svc = toolAccessPolicyService(db);
|
|
|
|
await expect(svc.createPolicy(company.id, {
|
|
name: "Unsafe regex",
|
|
policyType: "allow",
|
|
priority: 10,
|
|
selectors: { toolName: "send_email" },
|
|
conditions: {
|
|
arguments: { fieldMatches: { body: "^(a+)+$" } },
|
|
},
|
|
})).rejects.toThrow("unsafe regular expression");
|
|
});
|
|
|
|
it("rejects fieldMatches patterns with ambiguous alternation", async () => {
|
|
const company = await createCompany(db);
|
|
const svc = toolAccessPolicyService(db);
|
|
|
|
await expect(svc.createPolicy(company.id, {
|
|
name: "Unsafe alternation",
|
|
policyType: "allow",
|
|
conditions: {
|
|
arguments: { fieldMatches: { body: "^(a|aa)*b$" } },
|
|
},
|
|
})).rejects.toThrow("unsafe regular expression");
|
|
});
|
|
|
|
it("allows a write policy only for a safe argument subset and blocks unsafe arguments", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
const svc = toolAccessPolicyService(db);
|
|
const allowPolicy = await svc.createPolicy(company.id, {
|
|
name: "Allow safe destination",
|
|
policyType: "allow",
|
|
priority: 10,
|
|
selectors: { toolName: "send_email" },
|
|
conditions: {
|
|
arguments: {
|
|
fieldEquals: { to: "ops@example.com" },
|
|
fieldMatches: { body: "^[\\s\\S]{1,200}$" },
|
|
},
|
|
risk: { isWrite: true },
|
|
},
|
|
});
|
|
const blockPolicy = await svc.createPolicy(company.id, {
|
|
name: "Block external destination",
|
|
policyType: "block",
|
|
priority: 5,
|
|
selectors: { toolName: "send_email" },
|
|
conditions: {
|
|
arguments: {
|
|
fieldNotEquals: { to: "ops@example.com" },
|
|
},
|
|
},
|
|
});
|
|
|
|
await expect(svc.decide({
|
|
companyId: company.id,
|
|
actor: { actorType: "agent", actorId: agent.id, agentId: agent.id },
|
|
request: {
|
|
connectionId: connection.id,
|
|
catalogEntryId: catalogEntry.id,
|
|
toolName: "send_email",
|
|
arguments: { to: "ops@example.com", body: "safe update" },
|
|
},
|
|
})).resolves.toMatchObject({
|
|
allowed: true,
|
|
decision: "allow",
|
|
reasonCode: "allow_policy",
|
|
matchedPolicyIds: [allowPolicy.id],
|
|
policyExplanation: {
|
|
conditionsMatched: ["arguments", "risk"],
|
|
},
|
|
});
|
|
|
|
await expect(svc.decide({
|
|
companyId: company.id,
|
|
actor: { actorType: "agent", actorId: agent.id, agentId: agent.id },
|
|
request: {
|
|
connectionId: connection.id,
|
|
catalogEntryId: catalogEntry.id,
|
|
toolName: "send_email",
|
|
arguments: { to: "outside@example.com", body: "exfiltrate" },
|
|
},
|
|
})).resolves.toMatchObject({
|
|
allowed: false,
|
|
decision: "deny",
|
|
reasonCode: "deny_policy_block",
|
|
matchedPolicyIds: [blockPolicy.id],
|
|
});
|
|
});
|
|
|
|
it("fails closed for legacy unsupported redact and validate policies", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
const profile = await db.insert(toolProfiles).values({
|
|
companyId: company.id,
|
|
profileKey: `profile-${randomUUID()}`,
|
|
name: "Fallback allow",
|
|
defaultAction: "allow",
|
|
}).returning().then((rows) => rows[0]!);
|
|
await db.insert(toolProfileBindings).values(agentProfileBindingRow(company, profile, agent));
|
|
const [redactPolicy] = await db.insert(toolPolicies).values({
|
|
companyId: company.id,
|
|
name: "Legacy redact",
|
|
policyType: "redact" as never,
|
|
selectors: { toolName: "send_email" },
|
|
config: { redact: { fields: ["to", "body"] } },
|
|
priority: 1,
|
|
}).returning();
|
|
const [validatePolicy] = await db.insert(toolPolicies).values({
|
|
companyId: company.id,
|
|
name: "Legacy validate",
|
|
policyType: "validate" as never,
|
|
selectors: { toolName: "send_email" },
|
|
config: { schema: { required: ["body"] } },
|
|
priority: 2,
|
|
}).returning();
|
|
|
|
const redactDecision = await toolAccessPolicyService(db).decide(sendEmailDecisionInput(company, agent, connection, catalogEntry));
|
|
await db.update(toolPolicies).set({ enabled: false }).where(eq(toolPolicies.id, redactPolicy!.id));
|
|
const validateDecision = await toolAccessPolicyService(db).decide(sendEmailDecisionInput(company, agent, connection, catalogEntry));
|
|
|
|
expect(redactDecision).toMatchObject({
|
|
allowed: false,
|
|
decision: "deny",
|
|
reasonCode: "deny_policy_block",
|
|
matchedPolicyIds: [redactPolicy!.id],
|
|
});
|
|
expect(validateDecision).toMatchObject({
|
|
allowed: false,
|
|
decision: "deny",
|
|
reasonCode: "deny_policy_block",
|
|
matchedPolicyIds: [validatePolicy!.id],
|
|
});
|
|
});
|
|
|
|
it("fails closed for legacy condition-bearing policies", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
const profile = await db.insert(toolProfiles).values({
|
|
companyId: company.id,
|
|
profileKey: `profile-${randomUUID()}`,
|
|
name: "Fallback allow",
|
|
defaultAction: "allow",
|
|
}).returning().then((rows) => rows[0]!);
|
|
await db.insert(toolProfileBindings).values(agentProfileBindingRow(company, profile, agent));
|
|
const [policy] = await db.insert(toolPolicies).values({
|
|
companyId: company.id,
|
|
name: "Legacy conditional allow",
|
|
policyType: "allow",
|
|
selectors: { toolName: "send_email" },
|
|
conditions: { args: { body: "safe" } },
|
|
priority: 1,
|
|
}).returning();
|
|
|
|
const result = await toolAccessPolicyService(db).decide(sendEmailDecisionInput(company, agent, connection, catalogEntry));
|
|
|
|
expect(result).toMatchObject({
|
|
allowed: false,
|
|
decision: "deny",
|
|
reasonCode: "deny_policy_block",
|
|
matchedPolicyIds: [policy!.id],
|
|
});
|
|
});
|
|
|
|
it("fails closed for legacy rate-limit policies with invalid config", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
const profile = await db.insert(toolProfiles).values({
|
|
companyId: company.id,
|
|
profileKey: `profile-${randomUUID()}`,
|
|
name: "Fallback allow",
|
|
defaultAction: "allow",
|
|
}).returning().then((rows) => rows[0]!);
|
|
await db.insert(toolProfileBindings).values(agentProfileBindingRow(company, profile, agent));
|
|
const [policy] = await db.insert(toolPolicies).values({
|
|
companyId: company.id,
|
|
name: "Broken rate limit",
|
|
policyType: "rate_limit",
|
|
selectors: { toolName: "send_email" },
|
|
config: { rateLimit: { limit: 0, windowSeconds: 60 } },
|
|
priority: 1,
|
|
}).returning();
|
|
|
|
const result = await toolAccessPolicyService(db).decide(sendEmailDecisionInput(company, agent, connection, catalogEntry));
|
|
|
|
expect(result).toMatchObject({
|
|
allowed: false,
|
|
decision: "deny",
|
|
reasonCode: "deny_policy_block",
|
|
matchedPolicyIds: [policy!.id],
|
|
});
|
|
});
|
|
|
|
it("promotes repeated approved actions into a scoped trust rule with audited hits", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const issue = await createIssue(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
await db.insert(toolPolicies).values({
|
|
companyId: company.id,
|
|
name: "Review send_email",
|
|
policyType: "require_approval",
|
|
selectors: { toolName: "send_email" },
|
|
description: "Writes require review until promoted.",
|
|
});
|
|
const args = { to: "ops@example.com", body: "ship it" };
|
|
const first = await createApprovedToolAction({
|
|
db,
|
|
companyId: company.id,
|
|
agentId: agent.id,
|
|
connectionId: connection.id,
|
|
catalogEntryId: catalogEntry.id,
|
|
issueId: issue.id,
|
|
argumentsValue: args,
|
|
status: "executed",
|
|
});
|
|
await createApprovedToolAction({
|
|
db,
|
|
companyId: company.id,
|
|
agentId: agent.id,
|
|
connectionId: connection.id,
|
|
catalogEntryId: catalogEntry.id,
|
|
issueId: issue.id,
|
|
argumentsValue: args,
|
|
});
|
|
|
|
const trustRule = await toolAccessPolicyService(db).createTrustRuleFromActionRequest({
|
|
companyId: company.id,
|
|
actionRequestId: first.actionRequest.id,
|
|
body: {
|
|
approvalThreshold: 2,
|
|
scope: { includeIssue: true, includeCatalogEntry: true },
|
|
expiresAt: new Date(Date.now() + 60_000),
|
|
batchApproval: { enabled: true, maxBatchSize: 5, windowSeconds: 3600 },
|
|
},
|
|
});
|
|
const decision = await toolAccessPolicyService(db).decide({
|
|
...first.decisionInput,
|
|
consumeRateLimit: true,
|
|
});
|
|
const [updatedRule] = await db.select().from(toolPolicies).where(eq(toolPolicies.id, trustRule.id));
|
|
const trustConfig = updatedRule.config as { trustRule?: { hitCount?: number; lastHitAt?: string | null } };
|
|
const trustEvents = await db.select().from(toolCallEvents);
|
|
|
|
expect(trustRule).toMatchObject({
|
|
policyType: "trust_rule",
|
|
enabled: true,
|
|
selectors: expect.objectContaining({
|
|
agentId: agent.id,
|
|
issueId: issue.id,
|
|
connectionId: connection.id,
|
|
catalogEntryId: catalogEntry.id,
|
|
toolName: "send_email",
|
|
}),
|
|
});
|
|
expect(decision).toMatchObject({
|
|
allowed: true,
|
|
decision: "allow",
|
|
reasonCode: "allow_trust_rule",
|
|
matchedPolicyIds: [trustRule.id],
|
|
});
|
|
expect(trustConfig.trustRule?.hitCount).toBe(1);
|
|
expect(trustConfig.trustRule?.lastHitAt).toEqual(expect.any(String));
|
|
expect(trustEvents.some((event) => event.eventType === "trust_rule_created")).toBe(true);
|
|
expect(trustEvents.some((event) => event.eventType === "trust_rule_used")).toBe(true);
|
|
});
|
|
|
|
it("does not count approved actions outside the final trust-rule agent scope", async () => {
|
|
const company = await createCompany(db);
|
|
const agentA = await createAgent(db, company.id);
|
|
const agentB = await createAgent(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
await db.insert(toolPolicies).values(reviewSendPolicyRow(company, "Review scoped sends"));
|
|
const args = { to: "ops@example.com", body: "same reviewed payload" };
|
|
await createApprovedToolAction({
|
|
db,
|
|
companyId: company.id,
|
|
agentId: agentA.id,
|
|
connectionId: connection.id,
|
|
catalogEntryId: catalogEntry.id,
|
|
argumentsValue: args,
|
|
status: "executed",
|
|
});
|
|
const agentBAction = await createApprovedToolAction({
|
|
db,
|
|
companyId: company.id,
|
|
agentId: agentB.id,
|
|
connectionId: connection.id,
|
|
catalogEntryId: catalogEntry.id,
|
|
argumentsValue: args,
|
|
status: "executed",
|
|
});
|
|
|
|
await expect(toolAccessPolicyService(db).createTrustRuleFromActionRequest({
|
|
companyId: company.id,
|
|
actionRequestId: agentBAction.actionRequest.id,
|
|
body: { approvalThreshold: 2 },
|
|
})).rejects.toThrow(/final rule scope; found 1/);
|
|
});
|
|
|
|
it("does not count approvals from stale catalog versions toward trust-rule promotion", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
await db.insert(toolPolicies).values(reviewSendPolicyRow(company, "Review versioned sends"));
|
|
const args = { to: "ops@example.com", body: "same payload after tool change" };
|
|
await createApprovedToolAction(executedActionInput(db, company, agent, connection, catalogEntry, args));
|
|
await db
|
|
.update(toolCatalogEntries)
|
|
.set({ versionHash: randomUUID(), schemaHash: randomUUID(), updatedAt: new Date() })
|
|
.where(eq(toolCatalogEntries.id, catalogEntry.id));
|
|
const currentVersionAction = await createApprovedToolAction(executedActionInput(db, company, agent, connection, catalogEntry, args));
|
|
|
|
await expect(toolAccessPolicyService(db).createTrustRuleFromActionRequest({
|
|
companyId: company.id,
|
|
actionRequestId: currentVersionAction.actionRequest.id,
|
|
body: { approvalThreshold: 2, scope: { includeCatalogEntry: true } },
|
|
})).rejects.toThrow(/final rule scope; found 1/);
|
|
});
|
|
|
|
it("rejects company-wide trust-rule promotion bodies that drop the reviewed scope", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
await db.insert(toolPolicies).values(reviewSendPolicyRow(company, "Review scoped sends"));
|
|
const args = { to: "ops@example.com", body: "same reviewed payload" };
|
|
const first = await createApprovedToolAction(executedActionInput(db, company, agent, connection, catalogEntry, args));
|
|
await createApprovedToolAction(executedActionInput(db, company, agent, connection, catalogEntry, args));
|
|
|
|
await expect(toolAccessPolicyService(db).createTrustRuleFromActionRequest({
|
|
companyId: company.id,
|
|
actionRequestId: first.actionRequest.id,
|
|
body: {
|
|
approvalThreshold: 2,
|
|
scope: {
|
|
includeAgent: false,
|
|
includeProject: false,
|
|
includeApplication: false,
|
|
includeConnection: false,
|
|
includeTool: false,
|
|
},
|
|
},
|
|
})).rejects.toThrow(/reviewed actor\/tool scope/);
|
|
});
|
|
|
|
it("rejects argument-broadening trust-rule promotion bodies", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
await db.insert(toolPolicies).values(reviewSendPolicyRow(company, "Review exact payload sends"));
|
|
const args = { to: "ops@example.com", body: "same reviewed payload" };
|
|
const first = await createApprovedToolAction(executedActionInput(db, company, agent, connection, catalogEntry, args));
|
|
await createApprovedToolAction(executedActionInput(db, company, agent, connection, catalogEntry, args));
|
|
|
|
await expect(toolAccessPolicyService(db).createTrustRuleFromActionRequest({
|
|
companyId: company.id,
|
|
actionRequestId: first.actionRequest.id,
|
|
body: {
|
|
approvalThreshold: 2,
|
|
argumentFilters: { allowAny: true },
|
|
},
|
|
})).rejects.toThrow(/exact reviewed argument hash/);
|
|
});
|
|
|
|
it("falls back to review when a trusted catalog tool changes", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
await db.insert(toolPolicies).values(reviewSendPolicyRow(company, "Review changed sends"));
|
|
const args = { to: "ops@example.com", body: "repeatable" };
|
|
const first = await createApprovedToolAction(executedActionInput(db, company, agent, connection, catalogEntry, args));
|
|
await createApprovedToolAction({
|
|
db,
|
|
companyId: company.id,
|
|
agentId: agent.id,
|
|
connectionId: connection.id,
|
|
catalogEntryId: catalogEntry.id,
|
|
argumentsValue: args,
|
|
});
|
|
const trustRule = await toolAccessPolicyService(db).createTrustRuleFromActionRequest({
|
|
companyId: company.id,
|
|
actionRequestId: first.actionRequest.id,
|
|
body: { approvalThreshold: 2, scope: { includeCatalogEntry: true } },
|
|
});
|
|
await db
|
|
.update(toolCatalogEntries)
|
|
.set({ status: "quarantined", versionHash: randomUUID(), updatedAt: new Date() })
|
|
.where(eq(toolCatalogEntries.id, catalogEntry.id));
|
|
|
|
const decision = await toolAccessPolicyService(db).decide(first.decisionInput);
|
|
|
|
expect(decision).toMatchObject({
|
|
allowed: false,
|
|
decision: "require_approval",
|
|
reasonCode: "requires_review_changed_tool",
|
|
matchedPolicyIds: [trustRule.id],
|
|
});
|
|
});
|
|
|
|
it("revokes trust rules so matching actions return to per-call approval", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const { connection, catalogEntry } = await createTool(db, company.id);
|
|
await db.insert(toolPolicies).values(reviewSendPolicyRow(company, "Review revocable sends"));
|
|
const args = { to: "ops@example.com", body: "revocable" };
|
|
const first = await createApprovedToolAction(executedActionInput(db, company, agent, connection, catalogEntry, args));
|
|
await createApprovedToolAction({
|
|
db,
|
|
companyId: company.id,
|
|
agentId: agent.id,
|
|
connectionId: connection.id,
|
|
catalogEntryId: catalogEntry.id,
|
|
argumentsValue: args,
|
|
});
|
|
const trustRule = await toolAccessPolicyService(db).createTrustRuleFromActionRequest({
|
|
companyId: company.id,
|
|
actionRequestId: first.actionRequest.id,
|
|
body: { approvalThreshold: 2 },
|
|
});
|
|
const revoked = await toolAccessPolicyService(db).revokeTrustRule({
|
|
companyId: company.id,
|
|
policyId: trustRule.id,
|
|
body: { reason: "Tool scope changed" },
|
|
});
|
|
const decision = await toolAccessPolicyService(db).decide(first.decisionInput);
|
|
const config = revoked.config as { trustRule?: { revokedAt?: string | null; revocationReason?: string | null } };
|
|
|
|
expect(revoked.enabled).toBe(false);
|
|
expect(config.trustRule?.revokedAt).toEqual(expect.any(String));
|
|
expect(config.trustRule?.revocationReason).toBe("Tool scope changed");
|
|
expect(decision).toMatchObject({
|
|
allowed: false,
|
|
decision: "require_approval",
|
|
reasonCode: "requires_approval_policy",
|
|
});
|
|
});
|
|
|
|
it("routes gateway execution through policy decisions instead of legacy gateway permissions", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id, { toolGateway: { allowAll: true } });
|
|
const run = await createRun(db, company.id, agent.id);
|
|
const gateway = createToolGatewayService(db);
|
|
const session = await gateway.createSession({
|
|
companyId: company.id,
|
|
agentId: agent.id,
|
|
runId: run.id,
|
|
});
|
|
|
|
await expect(gateway.executeTool({
|
|
sessionToken: session.token,
|
|
tool: "mcp-remote-fixture:echo",
|
|
parameters: { message: "hello" },
|
|
})).rejects.toMatchObject({
|
|
status: 403,
|
|
reasonCode: "deny_default",
|
|
} satisfies Partial<ToolGatewayHttpError>);
|
|
|
|
const [invocation] = await db.select().from(toolInvocations);
|
|
expect(invocation).toMatchObject({
|
|
toolName: "mcp-remote-fixture:echo",
|
|
policyDecision: "deny",
|
|
status: "denied",
|
|
});
|
|
});
|
|
|
|
it("replays idempotent side-effecting gateway calls without creating a second invocation", async () => {
|
|
const company = await createCompany(db);
|
|
const agent = await createAgent(db, company.id);
|
|
const run = await createRun(db, company.id, agent.id);
|
|
await db.insert(principalPermissionGrants).values({
|
|
companyId: company.id,
|
|
principalType: "agent",
|
|
principalId: agent.id,
|
|
permissionKey: "tools:use",
|
|
scope: { toolName: "mcp-remote-fixture:update_note" },
|
|
});
|
|
const gateway = createToolGatewayService(db);
|
|
const session = await gateway.createSession({
|
|
companyId: company.id,
|
|
agentId: agent.id,
|
|
runId: run.id,
|
|
});
|
|
|
|
const first = await gateway.executeTool({
|
|
sessionToken: session.token,
|
|
tool: "mcp-remote-fixture:update_note",
|
|
parameters: { noteId: "n1", body: "ship" },
|
|
idempotencyKey: "note-update-1",
|
|
});
|
|
const replay = await gateway.executeTool({
|
|
sessionToken: session.token,
|
|
tool: "mcp-remote-fixture:update_note",
|
|
parameters: { noteId: "n1", body: "ship" },
|
|
idempotencyKey: "note-update-1",
|
|
});
|
|
|
|
expect(first).toMatchObject({ status: "completed", tool: "mcp-remote-fixture:update_note" });
|
|
expect(replay).toMatchObject({ status: "replayed", invocationId: first.invocationId });
|
|
const invocations = await db.select().from(toolInvocations);
|
|
expect(invocations).toHaveLength(1);
|
|
expect(invocations[0]).toMatchObject({
|
|
idempotencyKey: "note-update-1",
|
|
status: "succeeded",
|
|
resultSummary: expect.objectContaining({ summary: expect.any(String) }),
|
|
});
|
|
});
|
|
|
|
it("rejects cross-company owner agents and credential secret refs before persisting tool access records", async () => {
|
|
const company = await createCompany(db);
|
|
const otherCompany = await createCompany(db);
|
|
const otherAgent = await createAgent(db, otherCompany.id);
|
|
const [otherSecret] = await db.insert(companySecrets).values({
|
|
companyId: otherCompany.id,
|
|
key: `secret-${randomUUID()}`,
|
|
name: `Secret ${randomUUID()}`,
|
|
}).returning();
|
|
const svc = toolAccessService(db);
|
|
|
|
await expect(svc.createApplication(company.id, {
|
|
name: "Wrong owner",
|
|
type: "mcp_http",
|
|
ownerAgentId: otherAgent.id,
|
|
})).rejects.toThrow(/same company/);
|
|
|
|
await expect(svc.createConnection(company.id, {
|
|
name: "Wrong secret",
|
|
transport: "mcp_remote",
|
|
transportConfig: { url: "https://8.8.8.8/mcp" },
|
|
credentialSecretRefs: [{
|
|
secretId: otherSecret.id,
|
|
configPath: "headers.Authorization",
|
|
}],
|
|
})).rejects.toThrow(/same company/);
|
|
});
|
|
});
|