The agent tool landscape (optional)

Module 0: The agent tool landscape

Points: 10 · Type: orientation · Evidence: self-attested

Before the hands-on courses, get your bearings: what the tools are, what they have in common, and which one you will drive first.

The landscape

Six agentic coding tools carry this path’s hands-on courses. Pick by where you want to work:

Tool	Where it runs	What it’s good at
Claude Code	terminal	long multi-step tasks, strong instruction following, reads CLAUDE.md/AGENTS.md
Codex (desktop app)	desktop app	task threads with approvals and diff review, reads AGENTS.md
Cursor	IDE	Agent/Ask/Plan modes inside an editor, .cursor/rules project rules
DeepSeek Harness	Python framework	plugin-composed agents (cordis.yml), auditable append-only trajectories
Kimi Code	terminal	built-in explore/plan/coder subagents that split reading, planning, editing
ZCode	desktop app	long-horizon Goal Mode, task lists, workspace Git state, remote check-ins

They differ in interface, not in fundamentals. Every one of them: reads your files, follows written instructions, edits code, runs commands, and — this is the important part — can be wrong in ways that look right. That is why this path spends so much effort on rules and verification, not on prompts.

Choosing by environment: live in the terminal → Claude Code or Kimi Code. Live in a GUI → Codex app, Cursor, or ZCode. Want to assemble and audit the agent loop itself → DeepSeek Harness. You only need one to start; the route challenge at the end asks for a second, different one.

Your first task (no code)

1. Pick one of the six tools and install it following its official docs (linked from the course pages).
2. Open it on an empty folder and ask it: “create a Python script that prints the current date, then run it.”
3. Watch what it does: which files it creates, which commands it runs, what it reports back.
4. Ask it: “how do you know it works?” — notice whether it shows evidence or just says it worked. That gap is what the rest of this path is about.

Checkpoint

You have finished this module when you can answer, in your own words:

• Which agent did you install, and how do you start it?
• What did it do when you asked for the date script — and what evidence did it give you that the script ran?

Self-attested checkpoint: when you can answer both, record it — module M0 of the Foundation path is worth 10 points. Honest self-reporting is the rule of this platform: nobody checks your answer, but nothing here is a certificate.