CypherMonkeyAI omni-channel lab

A lifelong-learner lab on AI omni-channel agents, including voice and the contact center. Free to read. Nothing sold.

Prompting and evals · L1 · 15 min

A job, a shape, and a test

Learn in five minutes

  • A wish is not a job. Name the reader and the done check.
  • The shape is the output you will accept.
  • The test fails the change when the shape breaks.
  • One change at a time.
  • Keep the previous prompt next to the new one.

Replace 'be helpful' with 'return three bullets a new learner can check'. The second one can fail a test. The first one cannot.

This section HTML

Visual

T3-R1 diagram

Diagram SVGDiagram PNGDiagram HTML

Flow SVGFlow PNGFlow HTML

Rewrite one wish

  1. Start from 'make this better'.
  2. Write the reader: a learner opening the room.
  3. Write the shape: three bullets.
  4. Write the test: each bullet names a checkable act.

This section HTML

Quiz

What makes a prompt testable?

A test needs a shape.

How many changes should you score at once?

One change keeps the comparison fair.

What do you keep beside the new prompt?

The diff is the lesson.

This section HTML

Ecosystem

Done check
The condition that says the output is finished enough.

Sisters: JSON schema, Rubrics, Regression tests.

The same job-shape-test card works for chat, voice, and a tool result.

The capstone reuses this card.

Config

  • Write the goal in one sentence.
  • Name the input and the allowed output.
  • Save the trace next to the result.

Troubleshooting

  • The answer looks fluent and is wrong. Check whether a source passage actually contains the claim.
  • The loop never stops. Set a step cap and stop when the tool result repeats.

This section HTML

Sources HTML

Sources and go deeper

sources.json