{
  "action": "scaffold",
  "goal": "Migrate the billing module from callbacks to async/await",
  "max_iterations": 15,
  "protected_paths": [
    "tests/",
    "pyproject.toml"
  ],
  "tasks": [
    "Convert invoice.create to async",
    "Convert invoice.refund to async",
    {
      "size": "L",
      "title": "Update all call sites and then fix every failing test across the whole repo, and clean up"
    }
  ],
  "verify_commands": [
    {
      "command": "pytest -q tests/billing",
      "name": "unit tests"
    },
    {
      "command": "mypy billing/",
      "name": "types"
    }
  ]
}

{
  "action": "scaffold",
  "config": {
    "branch": "loop/migrate-the-billing-module-from-callback",
    "exit_conditions": [
      {
        "command": "pytest -q tests/billing",
        "expect_exit_code": 0,
        "name": "unit tests"
      },
      {
        "command": "mypy billing/",
        "expect_exit_code": 0,
        "name": "types"
      }
    ],
    "gate_commands": [
      {
        "command": "pytest -q tests/billing",
        "expect_exit_code": 0,
        "name": "unit tests"
      },
      {
        "command": "mypy billing/",
        "expect_exit_code": 0,
        "name": "types"
      }
    ],
    "goal": "Migrate the billing module from callbacks to async/await",
    "kit_dir": "loop-kit",
    "max_diff_lines": 400,
    "max_iterations": 15,
    "max_new_todo_per_iteration": 1,
    "max_total_minutes": 480,
    "protected_paths": [
      "tests/",
      "pyproject.toml"
    ],
    "recent_progress_entries": 5,
    "same_failure_limit": 3,
    "stall_limit": 3,
    "task_ids": [
      "T01",
      "T02",
      "T03"
    ],
    "timebox_minutes": 30
  },
  "files": [
    {
      "content": "# Loop prompt: Migrate the billing module from callbacks to async/await\n\nYou are ONE iteration of a bounded loop. You remember nothing from earlier iterations: the files in `loop-kit/` are your memory.\n\n## Read first (in this order)\n1. `loop-kit/todo.md`\n2. the last 5 entries of `loop-kit/progress.md`\n3. `loop-kit/EXIT_CONDITIONS.md`\n\n...",
      "description": "Per-iteration prompt the loop feeds to a fresh agent context.",
      "path": "loop-kit/PROMPT.md"
    },
    {
      "content": "# Todo: Migrate the billing module from callbacks to async/await\n\nOne iteration = one item. Keep each item small enough to finish and verify in a single iteration (sizes: S/M; split L).\n\n- [ ] T01 (S) Convert invoice.create to async\n- [ ] T02 (S) Convert invoice.refund to async\n- [ ] T03 (L) Update all call sites and then fix every failing test across the whole repo, and clean up",
      "description": "Ordered task list; exactly one unchecked item is worked per iteration.",
      "path": "loop-kit/todo.md"
    },
    {
      "content": "# Progress log (append-only)\n\nEntry template (copy for each iteration):\n\n```\n## Iteration N \u2014 YYYY-MM-DD HH:MM\n- item: \u003cid\u003e \u003ctitle\u003e\n- result: done | partial | failed\n- gate: \u003ccommand\u003e -\u003e exit \u003ccode\u003e (output tail: ...)\n...",
      "description": "Append-only iteration log; the agent\u0027s memory between iterations.",
      "path": "loop-kit/progress.md"
    }
  ],
  "findings": [
    {
      "code": "TASK_TOO_BIG",
      "message": "Task \u0027Update all call sites and then fix every failing test across\u0027 may not fit one iteration (size is L; bundles several actions; unbounded words (all/entire/everything)). Split it.",
      "severity": "warning",
      "where": "tasks[2]"
    }
  ],
  "lang": "en",
  "next_steps": [
    "Save the files and commit them on the loop branch.",
    "Dry-run one iteration before automating.",
    "After each iteration call the `evaluate` action."
  ],
  "playbook": "ralph-loop-playbook",
  "version": "1.0.0"
}

{
  "action": "lint_exit_conditions",
  "exit_conditions": [
    "make sure it works",
    "pytest -q || true",
    {
      "command": "npm test"
    }
  ]
}

{
  "action": "lint_exit_conditions",
  "findings": [
    {
      "code": "NOT_A_COMMAND",
      "message": "Exit condition \u0027make sure it works\u0027 is prose, not a runnable command. Replace it with a command whose exit code decides pass/fail.",
      "severity": "error",
      "where": "exit_conditions[0]"
    },
    {
      "code": "ALWAYS_PASSES",
      "message": "This command can never fail (\u0027|| true\u0027 / \u0027exit 0\u0027); it cannot gate completion.",
      "severity": "error",
      "where": "exit_conditions[1]"
    }
  ],
  "lang": "en",
  "passed": false,
  "playbook": "ralph-loop-playbook",
  "version": "1.0.0"
}

{
  "action": "fitness",
  "answers": {
    "decomposable": true,
    "safe_to_retry": true,
    "under_version_control": true,
    "verifiable_by_command": false
  }
}

{
  "action": "fitness",
  "advice": "Do not run an unattended loop. Fix the blockers first (add a verifying command, a branch, or a sandbox).",
  "answered": 4,
  "blockers": [
    {
      "message": "Required but false: Completion can be verified by a command (tests, build, lint, script) with a clear pass/fail.",
      "question": "verifiable_by_command"
    }
  ],
  "lang": "en",
  "missing": [
    "no_human_judgement",
    "has_baseline_tests",
    "bounded_scope"
  ],
  "playbook": "ralph-loop-playbook",
  "questions": [
    {
      "id": "verifiable_by_command",
      "required": true,
      "text": "Completion can be verified by a command (tests, build, lint, script) with a clear pass/fail."
    },
    {
      "id": "under_version_control",
      "required": true,
      "text": "Work happens in a git repository on a dedicated branch, so every iteration can be reverted."
    },
    {
      "id": "safe_to_retry",
      "required": true,
      "text": "A failed iteration has no irreversible side effects (no production writes, payments, emails, deletions)."
    },
    {
      "id": "decomposable",
      "required": false,
      "text": "The goal splits into small tasks that each fit in one iteration."
    },
    {
      "id": "no_human_judgement",
      "required": false,
      "text": "Done does not depend on subjective taste or stakeholder judgement."
    },
    {
      "id": "has_baseline_tests",
      "required": false,
      "text": "A passing baseline exists (existing tests or build) to detect regressions."
    },
    {
      "id": "bounded_scope",
      "required": false,
      "text": "Scope is bounded: the file/module area is known and changes stay inside it."
    }
  ],
  "score": 3,
  "verdict": "UNSUITABLE",
  "version": "1.0.0"
}

{
  "action": "evaluate",
  "history": [
    {
      "fingerprint": "TypeError:undefined",
      "gate": "fail",
      "items_done": 0
    },
    {
      "fingerprint": "TypeError:undefined",
      "gate": "fail",
      "items_done": 0
    },
    {
      "fingerprint": "TypeError:undefined",
      "gate": "fail",
      "items_done": 0
    }
  ]
}

{
  "action": "evaluate",
  "decision": "STOP_ESCALATE",
  "iteration": 3,
  "lang": "en",
  "message": "Stop the loop and escalate to a human with progress.md.",
  "playbook": "ralph-loop-playbook",
  "triggered": [
    {
      "action": "STOP",
      "message": "The same failure repeated 3 times.",
      "mode": "THRASH",
      "rule": "same_failure_limit"
    },
    {
      "action": "STOP",
      "message": "No item completed in 3 consecutive iterations.",
      "mode": "STALL",
      "rule": "stall_limit"
    }
  ],
  "version": "1.0.0"
}

{
  "action": "evaluate",
  "history": [
    {
      "gate": "pass",
      "items_done": 1
    },
    {
      "gate": "pass",
      "items_done": 1,
      "protected_changed": true
    }
  ]
}

{
  "action": "evaluate",
  "decision": "STOP_ESCALATE",
  "iteration": 2,
  "lang": "en",
  "message": "Stop the loop and escalate to a human with progress.md.",
  "playbook": "ralph-loop-playbook",
  "triggered": [
    {
      "action": "STOP",
      "message": "A protected path (tests/gate config) was modified. Halt for human review.",
      "mode": "FALSE_DONE",
      "rule": "protected_path_changed"
    }
  ],
  "version": "1.0.0"
}

{
  "action": "evaluate",
  "history": [
    {
      "exit_conditions_passed": true,
      "gate": "pass",
      "items_done": 1
    }
  ]
}

{
  "action": "evaluate",
  "decision": "DONE",
  "iteration": 1,
  "lang": "en",
  "message": "All exit-condition commands passed. Record the evidence and stop.",
  "playbook": "ralph-loop-playbook",
  "triggered": [],
  "version": "1.0.0"
}

{
  "action": "scaffold",
  "goal": "\uacb0\uc81c \ubaa8\ub4c8 \ube44\ub3d9\uae30 \uc804\ud658",
  "lang": "ko",
  "protected_paths": [
    "tests/"
  ],
  "tasks": [
    "invoice.create \ube44\ub3d9\uae30 \uc804\ud658"
  ],
  "verify_commands": [
    "pytest -q tests/billing"
  ]
}

{
  "action": "scaffold",
  "config": {
    "branch": "loop/work",
    "exit_conditions": [
      {
        "command": "pytest -q tests/billing",
        "expect_exit_code": 0,
        "name": "check-1"
      }
    ],
    "gate_commands": [
      {
        "command": "pytest -q tests/billing",
        "expect_exit_code": 0,
        "name": "check-1"
      }
    ],
    "goal": "\uacb0\uc81c \ubaa8\ub4c8 \ube44\ub3d9\uae30 \uc804\ud658",
    "kit_dir": "loop-kit",
    "max_diff_lines": 400,
    "max_iterations": 20,
    "max_new_todo_per_iteration": 1,
    "max_total_minutes": 480,
    "protected_paths": [
      "tests/"
    ],
    "recent_progress_entries": 5,
    "same_failure_limit": 3,
    "stall_limit": 3,
    "task_ids": [
      "T01"
    ],
    "timebox_minutes": 30
  },
  "files": [
    {
      "content": "# \ub8e8\ud504 \ud504\ub86c\ud504\ud2b8: \uacb0\uc81c \ubaa8\ub4c8 \ube44\ub3d9\uae30 \uc804\ud658\n\n\ub2f9\uc2e0\uc740 \uc81c\ud55c\ub41c \ub8e8\ud504\uc758 **\ud55c \ud68c\ucc28**\uc785\ub2c8\ub2e4. \uc774\uc804 \ud68c\ucc28\ub97c \uae30\uc5b5\ud558\uc9c0 \ubabb\ud558\uba70, `loop-kit/` \uc758 \ud30c\uc77c\uc774 \uae30\uc5b5\uc785\ub2c8\ub2e4.\n\n## \uba3c\uc800 \uc77d\uae30 (\uc21c\uc11c\ub300\ub85c)\n1. `loop-kit/todo.md`\n2. `loop-kit/progress.md` \uc758 \ucd5c\uadfc 5\uac1c \ud56d\ubaa9\n3. `loop-kit/EXIT_CONDITIONS.md`\n\n...",
      "description": "Per-iteration prompt the loop feeds to a fresh agent context.",
      "path": "loop-kit/PROMPT.md"
    },
    {
      "content": "# \ud560 \uc77c: \uacb0\uc81c \ubaa8\ub4c8 \ube44\ub3d9\uae30 \uc804\ud658\n\n\ud55c \ud68c\ucc28 = \ud55c \ud56d\ubaa9. \uac01 \ud56d\ubaa9\uc740 \ud55c \ud68c\ucc28\uc5d0 \ub05d\ub0b4\uace0 \uac80\uc99d\ud560 \uc218 \uc788\uc744 \ub9cc\ud07c \uc791\uc544\uc57c \ud569\ub2c8\ub2e4(\ud06c\uae30 S/M, L\uc740 \ucabc\uac1c\uae30).\n\n- [ ] T01 (S) invoice.create \ube44\ub3d9\uae30 \uc804\ud658",
      "description": "Ordered task list; exactly one unchecked item is worked per iteration.",
      "path": "loop-kit/todo.md"
    },
    {
      "content": "# \uc9c4\ud589 \uae30\ub85d (\ucd94\uac00 \uc804\uc6a9)\n\n\ud56d\ubaa9 \ud15c\ud50c\ub9bf(\ud68c\ucc28\ub9c8\ub2e4 \ubcf5\uc0ac):\n\n```\n## \ud68c\ucc28 N \u2014 YYYY-MM-DD HH:MM\n- \ud56d\ubaa9: \u003cid\u003e \u003c\uc81c\ubaa9\u003e\n- \uacb0\uacfc: done | partial | failed\n- \uac8c\uc774\ud2b8: \u003c\uba85\ub839\u003e -\u003e \uc885\ub8cc\ucf54\ub4dc \u003c\ucf54\ub4dc\u003e (\ucd9c\ub825 \ub05d\ubd80\ubd84: ...)\n...",
      "description": "Append-only iteration log; the agent\u0027s memory between iterations.",
      "path": "loop-kit/progress.md"
    }
  ],
  "findings": [],
  "lang": "ko",
  "next_steps": [
    "\ud30c\uc77c\uc744 \uc800\uc7a5\ud558\uace0 \ub8e8\ud504 \ube0c\ub79c\uce58\uc5d0 \ucee4\ubc0b.",
    "\uc790\ub3d9\ud654 \uc804\uc5d0 \ud55c \ud68c\ucc28\ub97c \uc2dc\ud5d8 \uc2e4\ud589.",
    "\ub9e4 \ud68c\ucc28 \ud6c4 `evaluate` \uc561\uc158 \ud638\ucd9c."
  ],
  "playbook": "ralph-loop-playbook",
  "version": "1.0.0"
}

{
  "action": "scaffold",
  "goal": "x",
  "verify_commands": [
    "pytest -q"
  ]
}

{
  "error": {
    "code": "INVALID_INPUT",
    "message": "\u0027tasks\u0027 must be an array with at least 1 item(s)."
  }
}