Agent Skills Workflow focused agent skills and evaluation suites for software development, skill maintenance, and Minecraft modding. The authoritative skill packages live under skills/ . Each package defines when the skill applies, what it may change, where it must stop, and what evidence it needs before claiming completion. Start Here For a multi turn coding workflow, explicitly invoke vibe coding with a concrete request: Codex: $vibe coding Claude Code: /vibe coding These are representative host syntaxes. vibe coding selects one visible specialist for the immediate phase; it does not run every phase at once or bypass the selected specialist's approval, write, verification, or stop rules. The examples assume the host can already see the skill. Host installation is environment specific; scripts/sync dev agent skills.py only manages this checkout's local .agents/skills/ snapshots and .claude/skills/ links. If you already know the task type, invoke the matching specialist directly. Use the tables below as a chooser, then read that skill's SKILL.md for the complete contract. For AI agents, this README is navigation rather than executable policy. Read the selected source SKILL.md completely and follow its required references before acting. Repository contributors must also follow AGENTS.md . Choose a Skill Workflow entry and coordination Task Skill Important boundary Package Route an explicitly invoked, multi turn coding workflow vibe coding Selects one primary visible specialist for the current phase and preserves that specialist's gates source · evals Confirm or correct the agent's understanding before ambiguous or risky work vibe goal alignment Produces an understanding record and stops before action until the user confirms or corrects it source · evals Coordinate bounded subagent research, edits, repairs, or review vibe orchestrate Provides delegation discipline; the coordinator keeps scope, verification, and consent ownership source · evals Shape, investigate, and