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.
uperLu's #50958 renamed plugins/cron → plugins/cron_providers but left
two test files patching the now-gone plugins.cron.chronos.verify path,
which would fail collection. Point them at plugins.cron_providers.*.
Add uperLu to release.py AUTHOR_MAP.
Live-test finding: the Chronos fire webhook was only on the APIServerAdapter
(aiohttp), but hosted agents expose `hermes dashboard` (the FastAPI web_server
app on :9119) as their public URL — NOT the api_server adapter. So NAS's relay
callback to {callback_url}/api/cron/fire could never reach the verifier on a
hosted agent (the exact target environment). Two layers were wrong:
1. Wrong server: /api/cron/fire didn't exist on the dashboard app. Added
cron_fire_webhook there, alongside the existing /api/cron/* dashboard routes.
It resolves the job's profile (_find_cron_job_profile) and runs fire_due via
the resolved provider under the cron-profile retarget lock
(_fire_cron_job_for_profile, mirroring _call_cron_for_profile) so the CAS
claim + run_one_job operate on the right profile's jobs.json. Runs with no
live adapters (delivery falls back to the per-platform send path, like the
desktop cron path). 202 + background so a long turn never trips NAS's
timeout; the store CAS de-dupes a NAS retry. job-not-found -> 200 "gone".
2. Auth gate: the dashboard auth middleware 401s any non-cookie request before
the handler runs. Added /api/cron/fire to the shared PUBLIC_API_PATHS so the
NAS bearer-JWT callback reaches the verifier — the JWT (purpose=cron_fire),
not the cookie, is the real gate. One shared frozenset feeds both the
loopback and OAuth middlewares, so no drift.
Kept the APIServerAdapter route too (valid self-host api_server surface).
Contract doc updated to name the dashboard app as the hosted-agent callback
surface.
Tests: test_cron_fire_dashboard (6) — route registered on the dashboard app,
in PUBLIC_API_PATHS, 401 on bad token WITH the cookie gate engaged (proves it's
reachable past the gate + JWT is the gate), 400 missing job_id, 200 gone for
unknown job, 202 + fire_due invoked for the resolved profile on a valid token.
Full hermes_cli + cron + chronos + webhook suites green (7637).
Why the original tests missed it: the api_server webhook test built an
APIServerAdapter client directly and never asserted which server the hosted
public URL exposes — green-but-wrong-integration. The new test pins the route
to the dashboard app.
