Skip to slide
166 / 170

CH 17 · Building the Full Mini-Harness · 07/11

Read as one page

Step 4: run it

# from the directory that contains the minharness/ package
python -m minharness.main /path/to/your/project

With the scripted call_model, you will see the agent take a request, ask to run ls (gated by the approval policy and the hook), feed the result back, and finish with a text message. Replace the body of call_model with a real API call and the same harness becomes a real coding agent: now the model decides what to run, and every safety, memory, context, and budget mechanism you built is already wrapped around it.

Finished this idea?

Your learning progress stays in this browser.