Export the trace as a JSON or text file, strip any student data, add a short plain-English note describing what you expected versus what happened, and share that package — not a screenshot.
Why Screenshots of Traces Are Useless to Developers
When something breaks in your campus AI agent, your first instinct might be to take a screenshot and send it to whoever is helping you fix it. Screenshots of traces are about as useful as photographing a car engine and asking a mechanic to diagnose it remotely. The mechanic needs to pop the hood, not squint at a photo.
A technical partner — whether that is a developer, a consultant, or an AI specialist — needs the raw trace data so they can search it, run queries against it, and reproduce the sequence of events. A screenshot captures one moment. A raw trace file captures everything.
How to Export and Prepare a Trace for Sharing
In Claude, you can copy the full conversation JSON from the API response or from the Claude.ai interface using the export option. In n8n, every workflow run has a JSON execution log you can download from the execution history panel. In Zapier, open the run history and use “View full details,” then copy the raw output. Save whatever you get as a .json or .txt file.
Before sharing, scan for student names, email addresses, or any personal data in the trace. Replace those with placeholders like [STUDENT_NAME] or [EMAIL_REDACTED]. This takes two minutes and keeps you on the right side of privacy rules.
Then write three sentences at the top of a separate note: what you asked the agent to do, what you expected it to do, and what it actually did instead. That context cuts your technical partner’s diagnosis time in half. They are not inside your head — they need to know the intended behaviour before they can identify the deviation.
What This Means for Educators
As a coach or course creator, you are probably not the person who built the underlying tools your agent uses. When something goes wrong — an agent sends the wrong email, pulls the wrong lesson, or gives a student bad information — you need to hand off the problem cleanly. A well-packaged trace with context and redacted student data is a professional handoff that gets problems solved faster.
It also protects you. If a student ever questions what your agent told them, having a clean trace export means you have a record of exactly what happened. That is true both for resolving disputes and for improving your agent over time.
What to Do Next
The next time your agent misbehaves, resist the screenshot reflex. Export the raw trace, redact personal data, write three sentences of context, and send that package. Most technical partners will have an answer for you within one conversation instead of five rounds of “can you send me more details.”