
Write, edit, refactor, or audit OpenClaw behavioral markdown so each surface does one job well and composes cleanly. Use when an OpenClaw agent needs stronger AGENTS.md, SOUL.md, TOOLS.md, IDENTITY.md, USER.md, MEMORY.md, HEARTBEAT.md, BOOTSTRAP.md, BOOT.md, SKILL.md, or HOOK.md authoring, layering, session-kind behavior, or memory layout.
Write, edit, refactor, or audit prompts so they stay intent-driven, section-correct, and anti-heuristic. Use when a prompt needs stronger commander's intent, cleaner section placement, safer refactoring without losing useful magic, or a findings-first audit.
Write, edit, refactor, or audit agent skills so they are leverage-first, self-contained, and anti-heuristic. Use when a Codex, OpenClaw, OpenAI, Anthropic, or repo-local skill needs stronger use-case selection, trigger boundaries, runtime-specific packaging, gating, references, or validation before shipping.
Control and interact with Kitty terminal windows - list windows, read output, send commands, and launch new terminals