Agent Builder
Watch the planner-executor loop light up, node by node.
The stop-condition stress test — five deliberately loop-inducing tasks — forces the classic failure mode ('agent loops and burns budget') to happen live, on camera, inside a hard ceiling instead of a real bill.
What goes in, what comes out
Learner defines an agent (persona, tools, budget, stop condition) and gives it a task; the planner-executor loop renders as a live state-machine graph that lights up node by node as the agent plans, calls a tool, observes, and re-plans. A visible step-budget ticker counts down in real time, and the lab includes five adversarial tasks specifically designed to induce infinite loops, so a weak stop condition fails visibly instead of silently burning spend.
A task description, an agent definition (persona + tools + budget + stop condition), and optionally a second agent for a handoff scenario.
A live state-machine trace of the planner-executor loop, a step-budget ticker, a pass/fail against the five adversarial loop-inducing tasks, and a final outcome log.
Agent persona/instructions, tool allowlist, step/token budget, stop-condition logic, single vs. multi-agent handoff, and approval-gate placement.
In-module lab for AIE-301/302 (SHOULD tier, core-agentic-credential module); appears as an observer mode in the Leadership specialization.
Go deeper, elsewhere
Hand-picked public explainers and open tools that complement this one — always optional, never required, never graded.
Concepts click when you open the machinery.
Three labs are already live and free — the same hands-on style this playground brings to its module.