[F003,F004,F014] enforce HMAC agent-token gate on do routes + WebSocket streams

F003/F014: /api/v1/do/* only required X-Agent-ID (UUID) — no token check,
unlike the flow routers' role guards. A forged X-Agent-ID passed. Added
require_any_authenticated_agent (token-only; do router serves all roles)
and applied it as a router-level dependency. Binds X-Agent-ID to a verified
HMAC token when ROBOCO_AGENT_AUTH_REQUIRED=true; rejects a forged token
even in dev mode.

F004: /ws/* per-agent streams (channels/agents/sessions/notifications)
never read the nginx-injected X-Agent-Token, so in strict mode an agent on
the Docker network could subscribe to another agent's notifications with
no auth. Added _require_panel_token verifying the CEO panel token against
the CEO identity; wired into all four per-agent streams (system stream
stays operator-only per its docstring). Same strict/dev contract.

TDD: RED tests watched fail (no gate -> 200/accept), then GREEN. ruff+mypy
clean; 399 api/mcp + 29 WS tests green, no regressions.
This commit is contained in:
Renn F
2026-06-28 10:04:55 +02:00
parent d579d3120a
commit 812bf1195a
5 changed files with 306 additions and 2 deletions
+35
View File
@@ -61,6 +61,41 @@ require_auditor = _require_roles(frozenset({Role.AUDITOR}))
require_pr_reviewer = _require_roles(frozenset({Role.PR_REVIEWER}))
def _require_authenticated_agent() -> params.Depends:
"""Token-only guard for the content-tool (do) router (F003/F014).
The do router serves every role — content tools are role-uniform, with
per-role removal handled in the spawn manifest — so, unlike the flow
routers, there is no single role to assert. But it must still bind the
presented ``X-Agent-ID`` to a verified HMAC token when
``ROBOCO_AGENT_AUTH_REQUIRED=true`` and reject a forged token even in
dev mode, exactly as the flow role guards do. Without this the
``/api/v1/do/*`` endpoints were the one agent-gateway path that
accepted a forged ``X-Agent-ID`` with no token check — a weaker gate
than ``/api/v1/flow/*``. The role/team headers are optional (the do
MCP server sends role but not team); they only feed the HMAC payload,
so a missing team is the empty-string team the token was issued with.
"""
def _check(
x_agent_id: Annotated[str, Header(alias="X-Agent-ID")],
x_agent_role: Annotated[str | None, Header(alias="X-Agent-Role")] = None,
x_agent_team: Annotated[str | None, Header(alias="X-Agent-Team")] = None,
x_agent_token: Annotated[str | None, Header(alias="X-Agent-Token")] = None,
) -> None:
from roboco.api.deps import _check_agent_auth_token
_check_agent_auth_token(
x_agent_id, x_agent_role or "", x_agent_team, x_agent_token
)
return cast("params.Depends", Depends(_check))
# The do router serves all roles, so this is token-only (no role assertion).
require_any_authenticated_agent = _require_authenticated_agent()
def envelope_to_response(env: Envelope, request: Request) -> dict[str, Any]:
"""Stamp the request's correlation_id onto the envelope and return wire-dict.
+12 -2
View File
@@ -6,7 +6,10 @@ from uuid import UUID
from fastapi import APIRouter, Depends, Header, Request
from roboco.api.deps import get_content_actions
from roboco.api.routes.v1._role_dep import envelope_to_response
from roboco.api.routes.v1._role_dep import (
envelope_to_response,
require_any_authenticated_agent,
)
from roboco.api.schemas.v1.do import (
ApprovePlaybookRequest,
ArchivePlaybookRequest,
@@ -31,7 +34,14 @@ from roboco.api.schemas.v1.do import (
)
from roboco.services.gateway.content_actions import ContentActions
router = APIRouter(prefix="/api/v1/do", tags=["v1-do"])
router = APIRouter(
prefix="/api/v1/do",
tags=["v1-do"],
# F003/F014: bind X-Agent-ID to a verified HMAC token — same gate the
# flow routers enforce via their role guards. The do router serves all
# roles, so this is token-only (no role assertion).
dependencies=[require_any_authenticated_agent],
)
_AgentIdHeader = Annotated[UUID, Header(alias="X-Agent-ID")]
_ContentActionsDep = Annotated[ContentActions, Depends(get_content_actions)]
+41
View File
@@ -21,6 +21,8 @@ from uuid import UUID
import httpx
from fastapi import APIRouter, WebSocket, WebSocketDisconnect, status
from roboco.agents_config import CEO_AGENT_ID, verify_agent_token
from roboco.api.deps import _auth_required
from roboco.config import settings
from roboco.db.base import get_db
from roboco.services.repositories import resolve_agent_uuid
@@ -28,6 +30,29 @@ from roboco.services.repositories import resolve_agent_uuid
router = APIRouter()
async def _require_panel_token(websocket: WebSocket) -> bool:
"""F004: bind a per-agent WS upgrade to the panel/CEO HMAC token.
The /ws/* streams are operator-only — the control panel is the sole WS
client (agents use MCP verbs, not WS), and nginx injects the CEO panel
token as ``X-Agent-Token`` on /ws/ upgrades. Without verifying it the
per-agent endpoints (channels/agents/sessions/notifications) accepted a
bare ``agent_id`` query param with no auth, so in strict mode
(``ROBOCO_AGENT_AUTH_REQUIRED=true``) an agent on the Docker network
could hit e.g. ``/ws/notifications/{id}`` directly and subscribe to
another agent's notifications. This gate requires + verifies the token
against the CEO identity in strict mode, and rejects a presented-but-
forged token even in dev mode — the same contract as the HTTP
``_check_agent_auth_token`` role gates. Returns True to proceed, False
to close with a policy violation (caller closes the socket).
"""
token = websocket.headers.get("x-agent-token")
if _auth_required() and not token:
return False
# A missing token in dev mode proceeds; a presented token must verify.
return not (token and not verify_agent_token(token, CEO_AGENT_ID, "ceo", ""))
# =============================================================================
# Connection Manager
# =============================================================================
@@ -262,6 +287,10 @@ async def channel_stream(
Clients receive real-time messages for the channel.
"""
# F004: verify the panel/CEO token before any subject lookup.
if not await _require_panel_token(websocket):
await websocket.close(code=status.WS_1008_POLICY_VIOLATION)
return
# Get agent ID from query params (or auth in production)
agent_id_str = websocket.query_params.get("agent_id")
if not agent_id_str:
@@ -318,6 +347,10 @@ async def agent_stream(
Clients receive real-time LLM output from the agent.
"""
# F004: verify the panel/CEO token before any subject lookup.
if not await _require_panel_token(websocket):
await websocket.close(code=status.WS_1008_POLICY_VIOLATION)
return
# Get viewer agent ID
viewer_id_str = websocket.query_params.get("viewer_id")
if not viewer_id_str:
@@ -365,6 +398,10 @@ async def session_stream(
Clients receive real-time messages for a specific session.
"""
# F004: verify the panel/CEO token before any subject lookup.
if not await _require_panel_token(websocket):
await websocket.close(code=status.WS_1008_POLICY_VIOLATION)
return
agent_id_str = websocket.query_params.get("agent_id")
if not agent_id_str:
await websocket.close(code=status.WS_1008_POLICY_VIOLATION)
@@ -410,6 +447,10 @@ async def notification_stream(
Agents receive real-time notifications via this stream.
"""
# F004: verify the panel/CEO token before any subject lookup.
if not await _require_panel_token(websocket):
await websocket.close(code=status.WS_1008_POLICY_VIOLATION)
return
# Validate agent exists in database
if not await validate_agent_exists(agent_id):
await websocket.close(code=status.WS_1008_POLICY_VIOLATION)
+97
View File
@@ -0,0 +1,97 @@
"""F003/F014: /api/v1/do/* must enforce the same HMAC agent-token gate as
the /api/v1/flow/* routers.
The do router serves every role (content tools are role-uniform), so it has
no single role to assert but it must still bind the presented X-Agent-ID
to a verified token when ROBOCO_AGENT_AUTH_REQUIRED=true and reject a forged
token even in dev mode. Without this guard the content-tool endpoints were
the one agent-gateway path that accepted a forged X-Agent-ID with no token
check a weaker gate than the flow routers' role guards.
"""
from __future__ import annotations
from typing import TYPE_CHECKING
from unittest.mock import AsyncMock, MagicMock
from fastapi import FastAPI
from fastapi.testclient import TestClient
from roboco.agents_config import issue_agent_token
from roboco.api.deps import get_content_actions
from roboco.api.routes.v1.do import router
from roboco.services.gateway.content_actions import ContentActions
if TYPE_CHECKING:
import pytest
_HTTP_200 = 200
_HTTP_401 = 401
_AGENT_ID = "00000000-0000-0000-0000-000000000001"
_SECRET = "test-secret-for-do-auth"
def _build_app() -> FastAPI:
app = FastAPI()
app.include_router(router)
mock_actions = MagicMock(spec=ContentActions)
mock_env = MagicMock()
mock_env.as_dict.return_value = {"status": "ok", "next": "continue"}
mock_actions.commit = AsyncMock(return_value=mock_env)
app.dependency_overrides[get_content_actions] = lambda: mock_actions
return app
def _commit_body() -> dict:
return {"message": "add user authentication endpoint"}
def test_do_route_401_when_auth_required_and_no_token(
monkeypatch: pytest.MonkeyPatch,
) -> None:
"""Strict mode: a do endpoint must require the token, not just X-Agent-ID."""
monkeypatch.setenv("ROBOCO_AGENT_AUTH_SECRET", _SECRET)
monkeypatch.setenv("ROBOCO_AGENT_AUTH_REQUIRED", "true")
client = TestClient(_build_app())
r = client.post(
"/api/v1/do/commit",
json=_commit_body(),
headers={"X-Agent-ID": _AGENT_ID, "X-Agent-Role": "developer"},
)
assert r.status_code == _HTTP_401
def test_do_route_rejects_forged_token_even_in_dev(
monkeypatch: pytest.MonkeyPatch,
) -> None:
"""Even in header-trust mode, a presented-but-forged token is rejected."""
monkeypatch.setenv("ROBOCO_AGENT_AUTH_SECRET", _SECRET)
monkeypatch.delenv("ROBOCO_AGENT_AUTH_REQUIRED", raising=False)
client = TestClient(_build_app())
r = client.post(
"/api/v1/do/commit",
json=_commit_body(),
headers={
"X-Agent-ID": _AGENT_ID,
"X-Agent-Role": "developer",
"X-Agent-Token": "forged-not-a-real-hmac",
},
)
assert r.status_code == _HTTP_401
def test_do_route_accepts_valid_token(monkeypatch: pytest.MonkeyPatch) -> None:
"""The good path: a valid HMAC token passes the guard and reaches the handler."""
monkeypatch.setenv("ROBOCO_AGENT_AUTH_SECRET", _SECRET)
monkeypatch.setenv("ROBOCO_AGENT_AUTH_REQUIRED", "true")
token = issue_agent_token(_AGENT_ID, "developer")
client = TestClient(_build_app())
r = client.post(
"/api/v1/do/commit",
json=_commit_body(),
headers={
"X-Agent-ID": _AGENT_ID,
"X-Agent-Role": "developer",
"X-Agent-Token": token,
},
)
assert r.status_code == _HTTP_200
+121
View File
@@ -0,0 +1,121 @@
"""F004: WebSocket streams must enforce the HMAC panel/CEO token gate when
ROBOCO_AGENT_AUTH_REQUIRED=true.
The /ws/* streams are operator-only (the panel is the sole WS client; agents
use MCP verbs, not WS). nginx injects the CEO panel token as X-Agent-Token on
/ws/ upgrades, but the endpoints never read or verified it so in strict mode
an agent on the Docker network could hit /ws/notifications/{id} directly and
subscribe to another agent's notifications with no auth. The fix binds each
per-agent WS upgrade to the CEO token: require + verify it in strict mode, and
reject a forged token even in dev mode (same contract as the HTTP role gates).
"""
from __future__ import annotations
from typing import TYPE_CHECKING
from unittest.mock import AsyncMock, MagicMock
from uuid import uuid4
import pytest
from fastapi import WebSocketDisconnect, status
from roboco.agents_config import CEO_AGENT_ID, issue_agent_token
from roboco.api.websocket import agent_stream, notification_stream
if TYPE_CHECKING:
import pytest as _pytest # noqa: F401
_SECRET = "test-secret-for-ws-auth"
def _mock_ws(headers: dict[str, str] | None, query: dict[str, str] | None) -> MagicMock:
ws = MagicMock()
ws.accept = AsyncMock()
ws.close = AsyncMock()
ws.send_json = AsyncMock()
ws.send_text = AsyncMock()
# One pong then disconnect so the receive loop exits after a successful gate.
ws.receive_text = AsyncMock(side_effect=["ping", WebSocketDisconnect()])
ws.headers = headers or {}
ws.query_params = query or {}
return ws
@pytest.mark.asyncio
async def test_notification_stream_rejects_missing_token_when_required(
monkeypatch: pytest.MonkeyPatch,
) -> None:
"""Strict mode + no X-Agent-Token => policy-violation close, never accepted."""
monkeypatch.setenv("ROBOCO_AGENT_AUTH_SECRET", _SECRET)
monkeypatch.setenv("ROBOCO_AGENT_AUTH_REQUIRED", "true")
agent_id = uuid4()
ws = _mock_ws(headers={}, query={})
with pytest.MonkeyPatch.context() as mp:
mp.setattr(
"roboco.api.websocket.validate_agent_exists",
AsyncMock(return_value=True),
)
await notification_stream(ws, agent_id)
ws.close.assert_awaited_once()
assert ws.close.await_args.kwargs["code"] == status.WS_1008_POLICY_VIOLATION
ws.accept.assert_not_awaited()
@pytest.mark.asyncio
async def test_notification_stream_rejects_forged_token_even_in_dev(
monkeypatch: pytest.MonkeyPatch,
) -> None:
"""Even in dev mode a presented-but-forged token is rejected."""
monkeypatch.setenv("ROBOCO_AGENT_AUTH_SECRET", _SECRET)
monkeypatch.delenv("ROBOCO_AGENT_AUTH_REQUIRED", raising=False)
agent_id = uuid4()
ws = _mock_ws(headers={"x-agent-token": "forged-not-a-real-hmac"}, query={})
with pytest.MonkeyPatch.context() as mp:
mp.setattr(
"roboco.api.websocket.validate_agent_exists",
AsyncMock(return_value=True),
)
await notification_stream(ws, agent_id)
ws.close.assert_awaited_once()
assert ws.close.await_args.kwargs["code"] == status.WS_1008_POLICY_VIOLATION
ws.accept.assert_not_awaited()
@pytest.mark.asyncio
async def test_notification_stream_accepts_valid_panel_token(
monkeypatch: pytest.MonkeyPatch,
) -> None:
"""A valid CEO panel token passes the gate and the socket is accepted."""
monkeypatch.setenv("ROBOCO_AGENT_AUTH_SECRET", _SECRET)
monkeypatch.setenv("ROBOCO_AGENT_AUTH_REQUIRED", "true")
token = issue_agent_token(CEO_AGENT_ID, "ceo", "")
agent_id = uuid4()
ws = _mock_ws(headers={"x-agent-token": token}, query={})
with pytest.MonkeyPatch.context() as mp:
mp.setattr(
"roboco.api.websocket.validate_agent_exists",
AsyncMock(return_value=True),
)
await notification_stream(ws, agent_id)
ws.accept.assert_awaited_once()
ws.close.assert_not_awaited()
@pytest.mark.asyncio
async def test_agent_stream_rejects_missing_token_when_required(
monkeypatch: pytest.MonkeyPatch,
) -> None:
"""The gate is wired into agent_stream too (viewer_id query param path)."""
monkeypatch.setenv("ROBOCO_AGENT_AUTH_SECRET", _SECRET)
monkeypatch.setenv("ROBOCO_AGENT_AUTH_REQUIRED", "true")
target_id = uuid4()
viewer_id = uuid4()
ws = _mock_ws(headers={}, query={"viewer_id": str(viewer_id)})
with pytest.MonkeyPatch.context() as mp:
mp.setattr(
"roboco.api.websocket.validate_agent_exists",
AsyncMock(return_value=True),
)
await agent_stream(ws, target_id)
ws.close.assert_awaited_once()
assert ws.close.await_args.kwargs["code"] == status.WS_1008_POLICY_VIOLATION
ws.accept.assert_not_awaited()