AgenticMail: Cross-agent task authorization bypass in AgenticMail APIGHSA-hjwc-26pj-v3pm

Summary

A low-privileged authenticated AgenticMail agent can enumerate another agent's pending/claimed tasks by supplying the target agent name to GET /api/agenticmail/tasks/pending?assignee=<name>. The returned task objects include the task IDs and payloads. The same task IDs can then be used with the capability-style task mutation endpoints (/tasks/:id/claim, /tasks/:id/result, /tasks/:id/complete, /tasks/:id/fail) to claim, complete, or fail tasks assigned to a different agent.

Because ordinary authenticated agents can discover agent names through GET /api/agenticmail/accounts/directory, the task ID effectively stops being a secret capability. This turns the intended capability model into a cross-agent authorization bypass.

Affected component

Package: @agenticmail/api Observed version: 0.9.62 Repository: agenticmail/agenticmail

Relevant code paths:

• packages/api/src/app.ts: createAuthMiddleware(...) is mounted before createAccountRoutes(...) and createTaskRoutes(...), so these routes are reachable by any valid bearer token.
• packages/api/src/routes/accounts.ts: GET /accounts/directory is available to any authenticated user and returns agent names.
• packages/api/src/routes/tasks.ts: GET /tasks/pending?assignee=name resolves arbitrary agent names and returns that agent's pending/claimed tasks.
• packages/api/src/routes/tasks.ts: /tasks/:id/claim, /tasks/:id/result, /tasks/:id/complete, /tasks/:id/fail, and /tasks/:id do not check whether the authenticated caller is the task assignee, assigner, or otherwise authorized for the task.

Impact

An attacker only needs a valid agent API key. They can:

1. List agent names using /accounts/directory.
2. Query another agent's task queue using /tasks/pending?assignee=<victimName>.
3. Read sensitive task payloads intended for the victim agent.
4. Use the disclosed task ID to complete/fail/claim the victim's task or submit attacker-controlled results.

Local reproduction

I reproduced this locally with a focused Vitest test mounted directly on createTaskRoutes. The test creates two agents, Alice and Bob, and one pending task assigned to Bob. Alice authenticates with her own agent key and performs the following sequence:

1. GET /api/agenticmail/tasks/pending?assignee=Bob with Authorization: Bearer ak_alice.
2. The response is HTTP 200 and includes Bob's task ID and payload: task-for-bob, { "task": "secret task intended for Bob" }.
3. Alice then sends POST /api/agenticmail/tasks/task-for-bob/complete with her own bearer token and an attacker-controlled result.
4. The task status becomes completed and the stored result is controlled by Alice.

The local verification command was:

npm run test --workspace=@agenticmail/api -- task-routes-authz.test.ts

Result:

PASS src/__tests__/task-routes-authz.test.ts (1 test)

Expected behavior

Task listing and task mutation endpoints should enforce an authorization relationship between the authenticated caller and the task. For example:

• GET /tasks/pending?assignee=<name> should either be restricted to the current agent, master/admin callers, or an explicit delegated relationship.
• /tasks/:id/claim, /tasks/:id/result, /tasks/:id/complete, /tasks/:id/fail, and /tasks/:id should verify that the caller is the assignee, assigner, master/admin, or otherwise explicitly authorized.
• If capability-based task IDs are retained, the API should not expose those IDs to unrelated agents through the assignee-name listing path.

Credit

Please credit the finder as: Yaohui Wang