Exam catalog

Clear goal, context, constraints, examples, iteration; reusing templates

Spotting hallucinations, subtle bugs and unsupported claims in seeded outputs; writing a test or check

Starting, directing, interrupting and reviewing a command-line AI agent; reading its diffs; approving or rejecting actions

Running several AI sessions in one terminal window using a multiplexer (tmux reference)

Saved prompts and snippets, scripts, scheduled jobs, connecting tools so repeated work is automatic

Not pasting secrets or customer data, handling prompt injection, license and IP awareness, review before commit or send

Choosing a suitable model size, managing context length, avoiding wasted runs

Matching each task to a suitable kind of AI tool and justifying the choice

Pointing AI at the right files, project instruction file, keeping secrets and noise out of context

Defining roles, naming every agent, isolating their work, handing off through files, merging results safely

A single timed 75-minute session: workspace, routing, context, team run, verify and merge, run report