fix: resolve lint errors from message-improvements

[ptone]

Summary

• Fix gofmt struct alignment in broker_v2.go after new field addition
• Fix errcheck: suppress unchecked json.NewEncoder(w).Encode() return in handlers.go
• Fix errcheck: suppress unchecked json.NewDecoder(rec.Body).Decode() returns in handlers_message_delivery_test.go (3 instances)

These lint fixes were in commit 0962ca8 on the message-improvements branch but were missed by the squash merge of PR GoogleCloudPlatform#409 (which used 526cdf4 as head).

Test plan

• CI passes (gofmt, golangci-lint, tests)

[gemini-code-assist]

Code Review

This pull request introduces external skill registries with new GCP and GitHub skill resolvers, adds a container image build command, transitions the legacy 'set[]' message syntax to 'group[]', and implements synchronous broker retries for message delivery. It also adds support for '--no-auth' agent startup with capture-auth staging, automatically cancels pending scheduled events upon agent deletion, and integrates OTel metrics for GCP tokens and broker authentication. The review feedback suggests parallelizing group message delivery to prevent timeouts, using reflection to dynamically set auth config fields, performing database updates asynchronously during agent deletion to reduce latency, and properly handling ignored errors when updating scheduled event statuses.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

Delivering group messages sequentially with a 30-second timeout per recipient can easily exceed the HTTP request timeout of the client or intermediate proxies (typically 60 seconds) if multiple recipients are offline or unreachable. This will cause a 504 Gateway Timeout for the entire request, failing delivery for subsequent recipients.

Since each recipient writes to a unique index in the results slice, there are no data races on the slice itself. Consider performing the dispatch in parallel using goroutines or errgroup to ensure that one offline agent does not block or timeout the entire group message request.

[gemini-code-assist]

The setAuthConfigField function uses a hardcoded switch statement to map string field names to api.AuthConfig struct fields. This introduces a maintenance burden whenever new auth fields are added. We can elegantly refactor this using Go's reflect package to dynamically set the fields by name, making the implementation completely generic and future-proof.

Note: You will need to import the "reflect" package in this file.

func setAuthConfigField(auth *api.AuthConfig, fieldName, value string) {
    v := reflect.ValueOf(auth).Elem()
    f := v.FieldByName(fieldName)
    if f.IsValid() && f.CanSet() && f.Kind() == reflect.String {
        f.SetString(value)
    }
}

[gemini-code-assist]

Running up to 1000 synchronous database updates in a loop inside the HTTP request path (performAgentDelete) can significantly increase latency and potentially cause gateway timeouts (e.g., 504 Gateway Timeout on Cloud Run/GCP). Consider performing these updates asynchronously in a background goroutine using a detached context (like context.WithoutCancel(ctx)), or utilizing a bulk update query if supported by the store.

[gemini-code-assist]

The error returned by UpdateScheduledEventStatus is silently ignored using _ = .... If the database update fails, the scheduled event might remain in a pending state or fail silently, making debugging extremely difficult. It is highly recommended to check the error and log it if it fails.

                if err := s.store.UpdateScheduledEventStatus(ctx, evt.ID, store.ScheduledEventFailed, &now, "target agent deleted"); err != nil {
                    slog.Error("Scheduler: failed to update scheduled event status", "eventID", evt.ID, "error", err)
                }