All labs · Lab 30 · intermediate · 45 min

A model plus a harness

Wrap your tool loop in the parts that make it usable: instructions in a file, a permission question, a budget and a log.

You build: harness.py: your loop from lab 13 that asks before it writes, stops at a budget, and logs every message.

  • Python 3.11 or newer and uv
  • An OpenRouter key (free models are enough)
  • Lab 13 done
OpenRouter docs

  1. 01Move the instructions into SYSTEM.md and load them at start

    uv run --with openai harness.py "What are your rules?"

    You pass when: It answers from SYSTEM.md. Change a rule in the file and the answer changes with no code edit.

  2. 02Add a write_file tool that asks y or n before it runs

    Reading is safe to allow. Writing, running commands and spending money ask first. That one split is most of agent safety.

    uv run --with openai harness.py "Create notes.txt with one line saying hello"

    You pass when: It asks before writing. Answer n: no file appears and the model is told it was refused.

  3. 03Stop the run at a turn limit and a token limit

    uv run --with openai harness.py --max-tokens 2000 "Read every file here and summarise each one"

    You pass when: It stops with a clear message that the budget ran out, and says how far it got.

  4. 04Append every message and tool result to a log file

    wc -l transcript.jsonl && tail -n 1 transcript.jsonl

    You pass when: One JSON line per message, with a timestamp and token counts. You can replay a run from the file alone.

The check runs on the computer, not on your word: a step passes when its command exits with 0 there.

Your computer · Fleet · off

Computers are not set up on this site yet. Run each command on your own machine and compare with "You pass when".