Track · Agent User
Agent tools foundation
Learn to drive real coding agents — Claude Code, Codex, Cursor, DeepSeek Harness, Kimi Code, or ZCode — give them good rules, and verify everything they produce.
TL;DR
Who it is forBeginners who have never (or barely) used AI coding agents.
What you finish withYou can independently drive two different coding agents on small tasks, write a single-source rules file, and verify agent output with your own hands.
The badgeAgent User — every required module claimed; a fork counts once you finish either side. Self-reported evidence, never a certificate.
The map
From start to badge
Follow the rail down. A fork is a pick-one choice — your agent, your call; both sides stay claimable as bonus practice.
Start here
Drop the skill link to your agent
One paste and it runs the whole track — the skill knows the loop.
Fork — pick one
- 06
One source of truth for agent rules
Course · 100 pts
- 07
Verifying AI-generated code
Course · 100 pts
- 09
The agent tool landscape (optional)
Orientation · 10 pts · optional — not required for the badge
Open the module → Finish
Badge: Agent User
Every required module claimed, capstone included — a fork needs either side, not both. Self-reported evidence, not a certificate.
Evidence
Points come from recorded claim codes
Checkpoint claims. python verify.py prints a code for passing test gates; self-reported checkpoints require --attest ID after the lesson questions. Each recorded checkpoint earns 10 points, a completed course badge adds 50, and the capstone earns 200.
Honest evidence. Objective checkpoints are gated by test suites; attested checkpoints are self-reported. Both are shown as self-reported evidence — never a certificate.