decide-cardlisted

# Decide a card Lean's **Andon cord** (Toyota production system): when a worker hits a problem they cannot resolve, they pull the cord; the line stops; a supervisor walks over, resolves the *cause*, and restarts the line. The system only works because both pulling AND lowering the cord are cheap. If lowering took fifteen minutes, workers would route around the cord and the whole quality system rots. Our cord is `human_gate`. `Skill(pull-card)` raises it (`none → decision` or `none → session`) when it hits a question only a human can answer — but only AFTER attempting the consuming repo's project-specific consultation (wired in via `.game-of-cards/hooks/decide-card.md`). `Skill(decide-card)` lowers the gate back to `none` in one action, which is what makes the line restart cheap. Status stays `open` so the next `pull-card` claims and implements per the recorded decision. The role split (lazy Andon): ``` HUMAN OR AGENT-WITH-RUBRIC AGENT decide what to do claim from gate=none queue record why implement lower the gate commit ``` Humans never wield the wrench; they only resolve the cause the worker flagged. Agents flag-and-wait when the project-specific rubric can't resolve the question — but agents that can cite the rubric MAY decide on their own behalf via this skill. The Andon-cord is now lazy: try the rubric, then pull. Keeps the human out of the loop when project-specific re