Assignments A1–A6
These six small assignments are the practical line through the first ten teaching weeks. Each is worth 2.5 points under the syllabus. Submit the artifact, verification evidence and a short explanation. End every report with decision: accept | narrow | recheck | stop, because: evidence IDs and next:.
Safety
Follow the safety practices in syllabus §9: run the agent in a dedicated directory or container; keep residency, employer and cloud credentials out of it; keep keys out of repositories, prompts, screenshots and logs; and run injection demonstrations offline against the local machine only. Use the mature coding-agent seat supplied for the course and do not run unknown generated commands on a credential-bearing host.
| Assignment | Week | Capability | Start |
|---|---|---|---|
| A1: Read a coding-agent session | 1 | Capture a real session and a forced failure | python practice/a1/check.py <submission> |
| A2: Build a small MCP tool server | 2 | Tool selection, arguments and recovery | python practice/a2/check.py <submission> |
| A3: Turn a manual workflow into a skill | 3 | Trigger and near-miss testing | python practice/a3/check.py <submission> |
| A4: Make a messy repository agent-ready | 5 | Instructions, hook and checks | python practice/a4/check.py <submission> |
| A5: Score agent readiness from observations | 6 | Anchored evidence and change conditions | python practice/a5/check.py <submission> |
| A6: Audit an evaluation instance and scorer | 10 | Denominators, counterexamples and limits | python practice/a6/check.py <submission> |
The checks provide public consistency feedback. They do not execute model-generated code, authenticate provenance or assign marks.