Tips // Claude in practice

Tips for working better with Claude

One short tip at a time, with an example you can copy. Move through them at random or search for what you need.

OtherIntermediateClaude

Runbook: symptom → architecture cause → first action

Teaching the team the path from symptom to cause lowers your future load; fixing the incident yourself only repeats it.

Symptoms group into a few causes: quality that drops with no code changes (model, prompt or retrieval drift), latency (context, a slow tool, a cache that does not hit), intermittent failures (auth, rate limits) and rising cost (model tier or cache). Add an escalation path that says what the team resolves and when it goes up.

Example
| Symptom | Likely cause | First action |
| Quality drops with no changes | model/prompt/drift | compare against the eval set |
| Cost rises with no more usage | tier or cache | check tier per request and hit rate |
#operations#runbook#support#observability