“Summarize what changed overnight: incidents, regressions, slow endpoints and open issues.”

Then posts a report every run

• Logs
• Traces
• Issues

“Compare this week with last week and name the three reliability fixes worth doing.”

Then posts a report every run

• Logs
• Traces

“Find errors that came back or grew after a quiet period.”

Then reports only when it finds something

• Logs

“Find endpoints whose p95 or p99 latency got materially worse than usual.”

Then reports only when it finds something

• Traces

“Flag services and endpoints whose traffic dropped or disappeared.”

Then reports only when it finds something

• Logs
• Traces

“Find downstream dependencies that got slower or started failing, and who calls them.”

Then reports only when it finds something

• Traces

“Compare this week’s user-facing errors and latency with last week’s.”

Then posts a report every run

• Logs
• Traces
• Metrics

“Find services that stopped sending logs, traces or metrics.”

Then reports only when it finds something

• Logs
• Traces
• Metrics

“Find missing or inconsistent service, environment and version attributes.”

Then reports only when it finds something

• Logs
• Traces

“Find logs and spans that can’t be joined because trace context is missing.”

Then reports only when it finds something

• Logs
• Traces

“Review each pull request against production evidence.”

Then posts a report every run

• Logs
• Traces
• Monitors

“List the services, routes and environments seen for the first time.”

Then reports only when it finds something

• Traces
• Endpoints

“Compare errors and latency before and after each new service version.”

Then reports only when it finds something

• Logs
• Traces
• Versions

“Investigate a Slack request and keep the evidence here.”

Then posts a report every run

• Logs
• Traces
• Metrics
• Issues

“Update new issues with a concise title, recommended action and severity assessment.”

Then posts a report every run

• Issues

“Write the on-call handoff: what’s open, what recovered and what to watch.”

Then posts a report every run

• Issues
• Incidents
• Logs
• Traces

“Check whether recent incidents are actually fixed in production.”

Then reports only when it finds something

• Incidents
• Issues

“Find noisy, stale or duplicate monitors, and the symptoms no monitor caught.”

Then reports only when it finds something

• Monitors
• Issues
• Incidents

“Find the services and messages producing the most avoidable log volume.”

Then reports only when it finds something

• Logs

“Catch sudden volume and cardinality growth before it shows up on the bill.”

Then reports only when it finds something

• Logs
• Traces
• Metrics

“Summarize unusual login and permission failures.”

Then reports only when it finds something

• Security logs