ax-agent-context

To manage context for long-context LLM agents across multiple runs, you need a recurring corpus orientation tool. This skill routes your requirement to contextMap for shared knowledge base retention and hands off to the appropriate memory skill.

The best way to compact agent history during a single run is within-run trajectory control. This skill identifies that need and routes you to the contextPolicy mechanism to manage evolving context windows effectively.

Use agent.optimize for improving behavior from examples and agent.playbook for evolving offline instructions. This skill distinguishes between these two offline tuning approaches and routes you to the correct optimization tool.

To choose the right Ax context tool for per-turn fact retrieval, you need a retrieval-based memory skill. This skill clarifies your turn-level lookup requirement and enforces handoff to the matching recall or discover skill.

Yes, you can use this routing skill to distinguish between contextMap and contextPolicy. It clarifies that contextMap applies to recurring corpus management while contextPolicy handles within-run trajectory control for long-context agents.

Your AxAgent workflow selects the wrong context management mechanism because routing logic is missing. This skill enforces appropriate handoff to ax-agent-rlm, ax-agent-optimize, or ax-agent-memory-skills to prevent incorrect tool selection.