AgentsGCPGCP Log Sink and Alert Review

GCP Log Sink and Alert Review

Review Cloud Logging sinks, log-based metrics, and alert policies for gaps in security monitoring coverage.

Tested against gcloud 4xx · cursor-0.45+ · windsurf-1.6+

DIRECTIVE / gcp-log-sink-alert-review
You are a GCP observability-security engineer. Your task is to review log sinks, log-based metrics, and alert policies and identify gaps in security monitoring coverage.

Context:
- You have read access to list log sinks, metrics, and alert policies.
- You will not modify any of them.

Steps:
1. List log sinks: gcloud logging sinks list, noting destination (BigQuery, Pub/Sub, Storage) and the filter expression; flag sinks with a narrow filter that may exclude security-critical log types.
2. Check log-based metrics: gcloud logging metrics list, and identify whether Security Command Center-important events (root login, bucket permissions change, firewall change) are represented by a metric.
3. Review alert policies: gcloud monitoring policies list and note which are enabled, their conditions, and their notification channels.
4. Map coverage: for a checklist of events (badge token exposure signs, critical project IAM change, sink deletion), confirm whether a metric exists.
5. Flag gaps: where a log type is not sink-ed, no metric tracks it, or no alert fires, record the missing piece.

Output format:
- Table: category, sink present, metric present, alert present, gap.
- A short prioritized list of the missing monitoring pieces.

Constraints:
- Read-only; never create sinks, metrics, or policies.
- Do not treat a raw log as coverage; an alert or metric must exist to make it actionable.
- Focus on the highest-value security events, not exhaustive coverage.

Example:
Input: a sink routes admin.activity to BigQuery, but no metric or alert exists for "DeleteBucket" or IAM policy changes.
Output: sink ok for admin.activity; gap: no metric/alert for IAM policy modifications -> recommend a log-based metric on protoPayload.methodName = google.iam.admin.v1.SetIamPolicy and a 5m alert.

Why this directive matters

Logging and alerting drift out of sync faster than either improves. A sink can quietly stop matching the events it was built for, or an alert can be defined on a query that no longer exists. This directive has the agent audit the three layers that monitoring actually depends on, sinks, log-based metrics, and policies, and reconcile them. It explicitly wants to know whether high-value security events like IAM policy changes and resource deletions are not only logged but turned into metrics that can fire alerts, because a raw log that nobody watches is not monitoring. The output is a coverage table per event category and a prioritized list of the missing pieces, so a team can see exactly where to add a metric or a policy. It is read-only and runs cleanly as a periodic check before or after a logging rearchitecture.

Test Command

gcloud logging sinks list && gcloud logging metrics list && gcloud monitoring policies list

Prerequisites

  • gcloud with logging and monitoring read
  • Project scope
  • Read-only access ok

Expected Outputs

  • Coverage table per event
  • Prioritized gap list
  • Suggested metrics/alerts

References

THE NEXT STEP

Automate it past the prompt.

This intermediate directive is a manual, read-only run. Devopsify can run the same check continuously across your estate, with policy gates, approvals, and a retained audit trail.