Author your own case

Build a broken system worth diagnosing.

How a Botopsy Lab diagnostic case is put together — a working system, a planted fault, the evidence to find it, and a grader that judges behavior. Build, break, diagnose, fix.

See live cases

Preview

Planned — not final. This documents the case-authoring flow we are building. The self-serve authoring tools are not open yet; the shape below is what we are working toward, not a shipped feature.

Anatomy of a case

Four parts, every time.

1. A working system, broken on purpose

Every case ships a real system that runs, then plants a fault in editable files. The learner opens files that actually break behavior — no toy stubs, no fill-in-the-blank.

2. Evidence commands to bound the fault

You script the commands a learner runs to see the failure and narrow it down — logs, node graphs, topic echoes, status checks. Diagnosis comes before editing.

3. A behavioral grader

The grader launches the learner's current files and judges observed runtime behavior — does the system now do the right thing? It is never a text diff against a reference answer.

4. A transfer challenge

A second, differently-shaped scenario that applies the same lesson elsewhere, so the skill sticks instead of memorizing one fix.

How you'll author one (planned)

From a fault idea to a gradable case.

Pick a fault class — the kind of failure you want to teach (a bad transform, a dropped message, a misconfigured lifecycle node).
Define the broken files plus a healthy reference — the editable broken state the learner starts from, and the working version that proves the fault is fixable.
Script the evidence — the commands and observations that let a learner bound the fault before touching code.
Write a behavioral grader — a pass/fail check that runs the learner's current files and reads observed behavior, not file contents.
Add a root-cause explanation and distractors — the real why, plus the plausible-but-wrong causes learners should rule out.
Add a transfer prompt — a related scenario that applies the same lesson in a new context.

Authored cases run in the same metered cloud workspace learners already use — the grader spins up a real container, not a sandboxed approximation.

What's live today vs coming

The cases are live. The builder is next.

Live now: 22 hand-built diagnostic cases and 9 browser playgrounds you can work through today. Each already runs on the behavioral-grader model described above.

Planned: the self-serve case builder that lets you author, grade, and publish your own case using this exact flow.

Work through a live case Join the community on Discord