Symptoms on unimplemented paths

Agent starts but sends nothing

The current agent composes configuration lifecycle, installation identity, and keys, but no collector or transport. Successful run is therefore not an upload test. Verify the source revision and agent status before investigating certificates. Use the backend development probe to test the implemented server path.

Agent spool is growing

There is no V2 spool implementation to diagnose at this revision. If you observe a real spool, identify whether it belongs to the legacy agent, a later V2 revision, or an external integration. Do not apply a fictional queue-path or purge command from the future architecture. For upcoming V2 implementations, compare oldest record age, byte bound, transient/permanent delivery failures, and durable ACK state as specified in the spool requirements.

Frontend cannot reach the API

The V2 frontend repository has no application yet. A running legacy UI is not automatically a compatible client for V2 protobuf endpoints or certificate/session authorization. Identify the client revision, route, encoding, and trust boundary. Do not add permissive CORS or a JSON adapter on the assumption that they are existing V2 configuration options.

Updates, policy, heartbeat, or commands do not work

These agent channels are planned and have no supported current endpoint/flag to turn on. Configuration explicitly refuses enabled updates. Backend agent liveness is read from telemetry, not from an agent heartbeat API. Refer to fleet-control target architecture for the actual requirements.

Verification

For a newly implemented revision, update the source baseline and confirm its composition root, contract, backend handler, and integration tests all exist before moving a feature to implemented status.

Source evidence

Reviewed against the source baseline. Seagull-agent-v2/cmd/seagull-agent/main.go · Seagull-agent-v2/internal/config/config.go · Seagull-frontend-v2/README.md · Seagull-backend-v2/internal/control/liveness.go.