future-agi/agent-learning-kit

future-agi/agent-learning-kit

Skill Claude CodeCodex

Builds the world an AI agent is tested in and everything its scenarios share, including the sub-goals they are checked on. Use after the agent has been understood and before any scenario is planned or written.

future-agi/agent-learning-kit

Skill Claude CodeCodex

Plans a suite of test scenarios for an AI agent and briefs the writers who write it. Use when a number of scenarios is asked for, before any is written, and whenever a round of writers has to be briefed or a finished suite checked before saving.

future-agi/agent-learning-kit

Skill Claude CodeCodex

Stands up the real service an AI agent connects to and proves it works, without touching the agent. Use when a test run needs the agent's own backing service running before scenarios are executed.

future-agi/agent-learning-kit

Skill Claude CodeCodex

Runs validated scenarios against an AI agent and explains what the results mean. Use once a suite has been saved and validated and the agent is reachable.

future-agi/agent-learning-kit

Skill Claude CodeCodex

Reads an AI agent's source or definition and records what is verifiably true about it: its tools, rules, data and use cases. Use first, before any world is built or scenario planned.

future-agi/agent-learning-kit

Skill Claude CodeCodex

Writes the scenarios an AI agent is tested with, each a whole task made hard in one real way and proved against the world before it is kept. Use whenever scenarios, test cases or a suite are wanted for an agent whose contract and world have already been built, and before every scenario is submitted.

future-agi/agent-learning-kit

Skill Claude CodeCodex

The one agent that builds an environment for an AI agent and owns it for the rest of its life. Knows how the whole system works and this environment end to end, answers from real data, builds, changes and repairs the environment through checked tools, runs it, and reports only what actually happened.

future-agi/agent-learning-kit

Skill Claude CodeCodex

What the platform supports around an environment - how it is built, what a scenario holds, how the Scenarios tab, coverage, runs, repeats and grading work, and what cannot be done. Use to answer any question about the platform or a limit, before promising anything.