autopilotOwn a feature, bugfix, or refactor end-to-end and return a finished, independently verified deliverable with evidence — not a chat reply. ALWAYS invoke when the user hands over a task to run without supervision: "autonomously build/ship/fix X", "take this end-to-end", "run this on autopilot", "I'm w
challengerBecome an adversarial Challenger that stress-tests a technical decision and its rationale — an architecture choice, a tech-stack pick, a design doc, a written spec, a migration plan — then forces a binary defend/revise verdict on the riskiest parts. Use this skill whenever someone presents reasoning
skill-forgeTurns the work you keep redoing by hand into a real, reusable Agent Skill — and proves it works before you trust it. ALWAYS invoke when the user wants to build, design, or scaffold a skill, agent, slash command, or CLAUDE.md harness; says "turn this workflow into a skill", "make a skill for X", "mak
system-organizerOrganize an agent system's knowledge, skills, commands, references and operational data. Use to audit a messy skill library, consolidate competing playbooks, retire obsolete context, reorganize an agent workspace, or extract maintenance rules from a working conversation. Supports review-only and aut