# PROMPTS — copy-paste prompts for learning with an agent

Paste these into any capable coding agent (Claude Code, Codex, etc.)
working in your kafka-lab directory, alongside this course pack. Each
prompt encodes the course's rules: hints first, evidence over opinions,
no invented results.

---

## 1. A small conceptual hint without the answer

```text
I'm working through "Kafka by Investigation", Exercise <NN>
(exercises/NN-<name>.md), section "<the bit I'm stuck on>". Give me ONE
small conceptual hint that points at the right idea without revealing the
answer. Do not mention the solution file. If my question itself contains a
wrong assumption, say so and stop there.
```

## 2. Reviewing my explanation

```text
Here is my written answer to the understanding-check question from
Exercise <NN>:

<paste your answer>

Read agent-guide/AGENTS.md first. Grade it against the exercise's
Expected observations and my actual outputs pasted below. Tell me which
parts are right, which are wrong, and for the wrong parts give me a hint
toward the fix — not the fix itself. Cite the exercise file and any lab
source files you read while grading.

My outputs: <paste the exact command output you observed>
```

## 3. Diagnosing a real error from my terminal

```text
I ran this, expecting <what you expected>:

$ <exact command>

and got this:

<paste the FULL terminal output, unedited>

Diagnose from evidence only: (1) restate what actually happened; (2) say
which parts of the output support each conclusion; (3) mark anything you
are unsure about as a hypothesis, not a finding; (4) point me at the
TROUBLESHOOTING.md section that matches, if one does. Do not change any
files or run anything destructive — suggest the next read-only command
for me to run instead.
```

## 4. Checking my result against the verification instructions

```text
Exercise <NN>'s Verification section says I'm done when:
<paste the exercise's verification lines>

Here is what I actually got:
<paste your outputs>

Check it honestly: does my evidence actually satisfy each item? Tell me
if anything is missing or if I'm claiming more than my output shows. Do
not mark anything verified that my output doesn't show. Read
VERIFICATION.md §<NN> and compare my observations with the recorded run,
noting (not fixing) any differences.
```

## 5. Prediction vs observed evidence

```text
Before running Exercise <NN> I predicted:
<paste your prediction>

After running it I observed:
<paste your output>

Explain the difference between what I predicted and what the broker
actually did. First, tell me whether my prediction was wrong about a
FACT (how Kafka behaves) or an ASSUMPTION (what state existed before I
ran). Then explain the actual mechanism, citing the official Kafka
docs for the config/behavior involved and naming the exact log lines or
columns in my output that prove it.
```

## 6. Reveal the solution — after I've tried

```text
I have finished my attempt at Exercise <NN> (my outputs and my answers
to the understanding checks are below). Now open solutions/<NN>-<name>.md
and grade my attempt against it: where did my explanation diverge, and
what did the recorded run show that I missed? Then, in one paragraph,
tell me the one thing I should carry into the next exercise.

My outputs: <paste>
My answers: <paste>
```

---

**One rule above all:** an agent that cannot see your real terminal output
is guessing. Paste the output, unedited — including the parts that look
like noise. The stacktrace is usually the lesson.
