✎ Text Section 1 of 1 draft

Content plan

Learners can set up an agent's working environment so it can act and self-correct.

Part 4 — Agentic execution

This part is dedicated to the AI-native craft. It is where the course title cashes out: orchestrating agents, in tactical terms, on top of everything built so far.

Outcome

Learners can set up an agent’s working environment so it can act and self-correct.

Planned coverage

  • The harness: everything the agent needs to do bounded work and check itself.
  • Tools and MCP: giving the agent the ability to act — run tests, query a database, hit an API — through tool use and MCP servers.
  • Tactical exposure: what to expose and what to withhold.
  • Skills and standards: encode the shared vocabulary and conventions from Module 4 as agent-readable artifacts.
  • Context discipline: hand the agent the bounded slice — the module or boundary — not the whole repository.

Agent payoff

A well-built harness is what turns a guessing agent into a self-correcting one.

Seeds

The harness is what the delivery plan’s tasks run inside.