hermes-agent/tests/tools/test_hermes_subprocess_env.py
Teknium 39975613b1
test: prune wave 2 + speed fixes — 28,106 → 19,757 test functions, suite wall 315s → 294s
Second, deeper pass over tools/gateway/hermes_cli plus first pass over
the trees wave 1 missed (acp, acp_adapter, skills, computer_use, docker,
dashboard, conformance, monitoring, secret_sources, hermes_state,
providers). Same rubric as wave 1 (AGENTS.md test policy); security,
alternation/caching invariants, issue-number regressions, and E2E kept.

Real test-quality fixes found and rooted out along the way:
- tests/tools/test_command_guards.py made real auxiliary-LLM HTTPS calls
  (DEFAULT_CONFIG smart-approval leaked in) — pinned approval
  mode=manual via autouse fixture: 17.4s → 0.4s.
- test_model_switch_custom_providers.py / test_user_providers_model_switch.py
  silently probed live provider catalogs (~2s/test) — stubbed
  cached_provider_model_ids/provider_model_ids/fetch_api_models.
- test_telegram_noise_filter.py: 15-platform copy-paste matrix over
  shared gateway.run logic → 3 representative platforms (55s → 3.9s).
- test_gateway_shutdown.py: stop()'s 5s interrupt-deadline loop spun on
  MagicMock agents — interrupt.side_effect now clears _running_agents
  (22s → 1.0s).
- test_gateway_inactivity_timeout.py poll-harness timings shrunk 3-5x
  (24s → 1.1s); test_mcp_stability.py backoff/SIGTERM-grace sleeps
  patched (15.4s → 2.5s); test_async_delegation.py negative-drain wait
  5s → 0.5s.
- test_telegram_init_deadline.py: loop-block margin restored to 1.0s
  with rationale comment — the watchdog-dump assertion needs the loop
  blocked well past deadline+grace under parallel load (flaked once in
  the 40-worker verification run at a 0.2s margin).

Verification: full hermetic suite via scripts/run_tests.sh —
2,438 files, 21,718 tests passed, 0 failed, 293.9s wall.
Suite totals vs original baseline: 46,820 → 19,757 test functions
(−57.8%), wall 583.5s → 293.9s (−50%), subprocess CPU 13,564s → 11,623s.
2026-07-29 13:39:40 -07:00

214 lines
8.1 KiB
Python

"""Tests for hermes_subprocess_env() — the centralized credential-safe env
builder for the non-terminal subprocess spawn surface.
Covers GHSA-m4m8-xjp4-5rmm / issue #29157: subprocesses spawned by the
gateway/browser/ACP/installer paths must not blindly inherit the operator's
full credential environment. Two tiers:
* Tier 1 (_ALWAYS_STRIP_KEYS): gateway bot tokens, GitHub auth, infra
secrets — stripped even when inherit_credentials=True.
* Tier 2 (_HERMES_PROVIDER_ENV_BLOCKLIST): LLM provider/tool keys — stripped
unless the caller opts into inherit_credentials=True.
"""
import os
from unittest.mock import patch
from tools.environments.local import (
hermes_subprocess_env,
_ALWAYS_STRIP_KEYS,
_HERMES_PROVIDER_ENV_FORCE_PREFIX,
)
_TIER1_SAMPLE = {
"GH_TOKEN": "ghp_secret",
"TELEGRAM_BOT_TOKEN": "bot-token",
"SLACK_APP_TOKEN": "xapp-secret",
"MODAL_TOKEN_SECRET": "modal-secret",
"HERMES_DASHBOARD_SESSION_TOKEN": "dash-secret",
}
_PROVIDER_SAMPLE = {
"OPENAI_API_KEY": "sk-fake",
"ANTHROPIC_API_KEY": "ant-fake",
"OPENROUTER_API_KEY": "or-fake",
}
_SAFE_SAMPLE = {
"PATH": "/usr/bin:/bin",
"HOME": "/home/user",
"USER": "testuser",
"MY_APP_VAR": "keep-me",
}
def _build(extra=None, *, inherit_credentials=False):
env = dict(_SAFE_SAMPLE)
if extra:
env.update(extra)
with patch.dict(os.environ, env, clear=True):
return hermes_subprocess_env(inherit_credentials=inherit_credentials)
class TestStripByDefault:
def test_provider_keys_stripped_by_default(self):
result = _build(_PROVIDER_SAMPLE)
for var in _PROVIDER_SAMPLE:
assert var not in result, f"{var} leaked with inherit_credentials=False"
def test_tier1_secrets_stripped_by_default(self):
result = _build(_TIER1_SAMPLE)
for var in _TIER1_SAMPLE:
assert var not in result, f"{var} leaked (Tier-1) with inherit_credentials=False"
def test_pythonutf8_set(self):
result = _build()
assert result.get("PYTHONUTF8") == "1"
class TestInheritCredentials:
def test_provider_keys_preserved_when_inheriting(self):
result = _build(_PROVIDER_SAMPLE, inherit_credentials=True)
for var, val in _PROVIDER_SAMPLE.items():
assert result.get(var) == val, f"{var} should survive inherit_credentials=True"
def test_tier1_secrets_stripped_even_when_inheriting(self):
"""The whole point of Tier 1: gateway/GitHub/infra secrets never reach
a child, even a model-driving CLI that legitimately needs provider keys."""
result = _build({**_PROVIDER_SAMPLE, **_TIER1_SAMPLE}, inherit_credentials=True)
for var in _TIER1_SAMPLE:
assert var not in result, (
f"{var} (Tier-1) must be stripped even with inherit_credentials=True"
)
# ...while provider keys survive.
for var in _PROVIDER_SAMPLE:
assert var in result
def test_pythonutf8_set_when_inheriting(self):
assert _build(inherit_credentials=True).get("PYTHONUTF8") == "1"
class TestTierInvariants:
def test_tier1_always_stripped_both_paths(self):
"""Behavioral invariant: every Tier-1 key is stripped on BOTH the
default path and the inherit_credentials=True path. This is what
guarantees no gap, regardless of whether the key also happens to be
in the provider blocklist."""
sample = {k: f"secret-{k}" for k in _ALWAYS_STRIP_KEYS}
for inherit in (False, True):
result = _build(sample, inherit_credentials=inherit)
leaked = {k for k in _ALWAYS_STRIP_KEYS if k in result}
assert not leaked, (
f"Tier-1 keys leaked with inherit_credentials={inherit}: {sorted(leaked)}"
)
def test_tier1_covers_gateway_bot_token(self):
assert "TELEGRAM_BOT_TOKEN" in _ALWAYS_STRIP_KEYS
def test_tier1_covers_github_auth(self):
assert {"GH_TOKEN", "GITHUB_TOKEN"} <= _ALWAYS_STRIP_KEYS
def test_tier1_covers_infra_secrets(self):
assert {"MODAL_TOKEN_ID", "MODAL_TOKEN_SECRET", "DAYTONA_API_KEY"} <= _ALWAYS_STRIP_KEYS
class TestBrowserPassthroughPattern:
def test_browser_keys_recoverable_after_strip(self):
"""Browser tool pattern: strip everything, then re-add the browser
backend keys agent-browser actually needs."""
from tools.browser_tool import _BROWSER_PASSTHROUGH_KEYS
leaked = {
"BROWSERBASE_API_KEY": "bb-key",
"BROWSERBASE_PROJECT_ID": "bb-proj",
"FIRECRAWL_API_KEY": "fc-key",
"ANTHROPIC_API_KEY": "ant-should-go",
"TELEGRAM_BOT_TOKEN": "bot-should-go",
}
with patch.dict(os.environ, {**_SAFE_SAMPLE, **leaked}, clear=True):
env = hermes_subprocess_env(inherit_credentials=False)
for key in _BROWSER_PASSTHROUGH_KEYS:
if key in os.environ:
env[key] = os.environ[key]
assert env["BROWSERBASE_API_KEY"] == "bb-key"
assert env["FIRECRAWL_API_KEY"] == "fc-key"
# Provider + gateway secrets must NOT come back.
assert "ANTHROPIC_API_KEY" not in env
assert "TELEGRAM_BOT_TOKEN" not in env
class TestDelegatedChildMarker:
def test_delegated_child_context_scrubs_parent_kanban_keys_and_sets_marker(self):
from agent.delegation_context import delegated_child_context
with patch.dict(
os.environ,
{
**_SAFE_SAMPLE,
"HERMES_KANBAN_TASK": "t_parent",
"HERMES_KANBAN_RUN_ID": "123",
"HERMES_KANBAN_DB": "/tmp/parent-kanban.db",
"HERMES_KANBAN_WORKSPACE": "/tmp/parent-workspace",
},
clear=True,
):
with delegated_child_context():
env = hermes_subprocess_env(inherit_credentials=True)
assert env["HERMES_DELEGATED_CHILD_CONTEXT"] == "1"
assert "HERMES_KANBAN_TASK" not in env
assert "HERMES_KANBAN_RUN_ID" not in env
assert "HERMES_KANBAN_DB" not in env
assert "HERMES_KANBAN_WORKSPACE" not in env
assert env["MY_APP_VAR"] == "keep-me"
_INTERNAL_DYNAMIC_SAMPLE = {
"AUXILIARY_VISION_API_KEY": "sk-vision",
"AUXILIARY_VISION_BASE_URL": "http://internal:1234/v1",
"AUXILIARY_WEB_EXTRACT_API_KEY": "sk-webx",
"GATEWAY_RELAY_SECRET": "relay-secret",
"GATEWAY_RELAY_DELIVERY_KEY": "relay-delivery",
}
class TestInternalDynamicSecrets:
"""AUXILIARY_*_API_KEY / _BASE_URL and GATEWAY_RELAY_* auth are stripped on
BOTH paths — including inherit_credentials=True — since a model-driving CLI
(codex/copilot) never needs them even when it needs provider keys."""
def test_stripped_by_default(self):
result = _build(_INTERNAL_DYNAMIC_SAMPLE)
for var in _INTERNAL_DYNAMIC_SAMPLE:
assert var not in result, f"{var} leaked with inherit_credentials=False"
def test_auxiliary_non_secrets_preserved(self):
"""AUXILIARY_*_PROVIDER / _MODEL routing config survives (not secrets)."""
result = _build(
{"AUXILIARY_VISION_PROVIDER": "openai", "AUXILIARY_VISION_MODEL": "gpt-4o"},
)
assert result.get("AUXILIARY_VISION_PROVIDER") == "openai"
assert result.get("AUXILIARY_VISION_MODEL") == "gpt-4o"
def test_gateway_relay_id_stripped_even_when_inheriting(self):
"""GATEWAY_RELAY_ID has no secret suffix (predicate skips it) but is
gateway-identifying auth material provisioned alongside the relay
secret. It's in _ALWAYS_STRIP_KEYS so it's stripped on the inherit path
too — closes the codex/copilot leak the predicate alone would miss."""
result = _build(
{**_PROVIDER_SAMPLE, "GATEWAY_RELAY_ID": "relay-id"},
inherit_credentials=True,
)
assert "GATEWAY_RELAY_ID" not in result
# provider keys still flow (codex auth)
for var in _PROVIDER_SAMPLE:
assert var in result
def test_relay_triplet_in_always_strip(self):
assert {
"GATEWAY_RELAY_ID", "GATEWAY_RELAY_SECRET", "GATEWAY_RELAY_DELIVERY_KEY",
} <= _ALWAYS_STRIP_KEYS