Case study
An AI skill for on-call debugging
An AI skill that turns an error link into a likely cause, a proposed fix and a Slack update, now part of on-call training.
Context
On-call engineers investigate alerts by digging through logs in Datadog and Sentry, often under time pressure.
Challenge
Investigations can be slow and stressful, especially for engineers who are new to on-call.
Approach
Built a skill for Claude and Cursor. Given a link to an error in Datadog or Sentry, it:
- Reads the logs.
- Finds the likely cause of the issue.
- Proposes a solution.
- Shares the results with the team on Slack.
Outcome
The skill became part of the on-call training program. New engineers can see how an investigation unfolds, step by step, before they face a real alert alone.
What I learned
The most valuable AI tools don’t replace engineers. They take a stressful moment and make it calmer.