SCENARIO 1
The missing module
A Node.js app says Cannot find module 'express'. The AI recommends upgrading Node and rewriting imports.
AI debugging training ground
Diagnose the prompt before asking AI to diagnose the code
Measure whether a debugging request contains the evidence needed for a useful answer, generate a clearer structure, and practice choosing the next move in realistic failure scenarios.
CLINIC RULE
A longer prompt is not automatically better. Every added line should locate the failure, explain causality, or rule out a hypothesis.
Prompt checkup
Paste a debugging request and find the missing evidence
The rubric is deterministic and runs in this browser. It evaluates prompt structure, not whether a proposed diagnosis is correct.
Prompt readiness
0
OUT OF 100
Failure action
MISSING
Add the command, route, test, or user action that fails.
Environment
MISSING
Name the runtime, framework, operating system, and relevant versions.
Expected behavior
MISSING
State what should happen and how the actual result differs.
Error evidence
MISSING
Include a reviewed error excerpt, assertion difference, or exit code.
Recent change
MISSING
Describe the most relevant code, dependency, or configuration change.
Checks already completed
MISSING
List the checks already completed so the answer does not repeat them.
Requested outcome
MISSING
Ask for one outcome: root cause, next check, minimal fix, or reproduction plan.
Suggested structure
Please identify the likely root cause and the next diagnostic step. Command or failing action: [add this] Environment and versions: [add this] Expected behavior: [add this] Recent relevant change: [add this] Checks already completed: [add this] Reviewed and redacted error: [paste the relevant evidence]
Evidence challenge
Choose the next move in five debugging situations
Each answer explains why a tempting shortcut is weaker than an evidence-driven next step.
SCENARIO 2
The hydration mismatch
The server renders June 8 while the first client render shows June 7. The AI suggests disabling SSR.
SCENARIO 3
The failing CI test
Checkout and install succeed. One test expects 200, receives 500, and throws in src/user.ts.
SCENARIO 4
The wrapped Python error
InvalidUser is raised from a ValueError. The AI discusses only InvalidUser.
SCENARIO 5
The confident patch
The AI proposes changes in four files but never states the root cause or a verification command.
PROGRESS 0/5
0 correct so far
Understand the cleaner
Review the exact filtering rules and known failure modes.
Study annotated cases
See why specific lines stay, leave, or need to be restored.
Recognize AI failure patterns
Learn the habits that make confident answers unreliable.