PlainLogic Arcade · AI Skills
Agent Mission
You are the planner for an AI agent. Each mission hands you a goal and a toolbox — pick the right tools in the right order, then watch the agent loop run: goal, reasoning, tool, result, answer.
The game
Run the agent
Play · Watch · Learn
How it works
1 · Pick the tools
Each mission gives you a goal and a toolbox. Click tools in the order the agent should use them — sometimes the answer is one tool, sometimes two, sometimes no tool at all.
2 · Watch the loop run
Hit Run the agent and watch the loop play out live: goal → reasoning → tool → result → answer. The reasoning steps are simplified illustrations, not real model internals.
3 · Read the why
Every mission ends with the lesson it taught: sometimes rules win, sometimes data wins, and sometimes a human should decide. Five missions, three points each.
Under the hood
What this teaches
An agent is not a smarter chatbot. A chatbot turns your words into more words. An agent runs a loop: it takes a goal, does some reasoning about what it needs, calls tools to get it, reads the result, and repeats until it can answer. The loop — not the cleverness — is the whole idea.
Picking tools is the skill. The model's job is reasoning; your job (or the agent designer's job) is making sure the right tools exist and get used in the right order. A weather question needs a weather tool, not a bigger brain. Three prices need a calculator, not a web search.
Knowing when to stop is a superpower. The best tool call is sometimes no tool call, and the bravest thing an agent can do is admit no tool fits and hand the decision to a human. Confidence without capability is just lying with extra steps.