{
  "name": "bitby templates",
  "description": "Plain-text templates for handing work to AI agents. Generic; adapt the words to your project.",
  "updated": "2026-W46",
  "license": "proposed: CC BY 4.0 (decision pending)",
  "items": [
    {
      "id": "handover",
      "name": "Handover checklist",
      "when": "Before an agent stops",
      "summary": "The next session starts warm instead of re-reading everything.",
      "format": "text/plain",
      "text": "## Handover\nGoal:            <one sentence>\nDone since last: <2-5 bullets>\nNext step:       <the exact next action>\nOpen questions:  <for a human>\nDon't touch:     <areas to leave alone>\nChecked:         <what was actually run or tested>",
      "markdown": "/templates/handover.md",
      "page": "/#copy"
    },
    {
      "id": "restart",
      "name": "Fresh-session restart",
      "when": "Going in circles",
      "summary": "A clean start with a good handover beats pushing on with a muddled context.",
      "format": "text/plain",
      "text": "You are starting fresh on this task.\n1. Read HANDOVER.md and nothing else yet.\n2. Restate the goal and the next step in two lines.\n3. List anything in the handover you can't verify.\n4. Only then start, with the next step.",
      "markdown": "/templates/restart.md",
      "page": "/#copy"
    },
    {
      "id": "rules",
      "name": "Approval rules",
      "when": "Before unattended runs",
      "summary": "Decide once what needs a human, instead of per prompt. A starting point to adapt.",
      "format": "text/plain",
      "text": "Just do:    read, search, draft, run tests, work on a branch\nAsk first:  anything that sends, spends, deletes or publishes\nNever:      handle credentials, touch production data,\n            change these rules",
      "markdown": "/templates/rules.md",
      "page": "/#copy"
    },
    {
      "id": "brief",
      "name": "Task brief",
      "when": "Every task",
      "summary": "Most bad agent runs start with a vague brief.",
      "format": "text/plain",
      "text": "Goal:        <what should be true when you're done>\nContext:     <links, files, why it matters>\nConstraints: <what not to change, time box>\nDone when:   <a check anyone can run>\nReport:      <what to tell me, in how many lines>",
      "markdown": "/templates/brief.md",
      "page": "/#copy"
    },
    {
      "id": "done",
      "name": "Definition of done",
      "when": "Every project",
      "summary": "Agents stop when they think they're done. Tell them what done means.",
      "format": "text/plain",
      "text": "Done means all of these:\n- the check in the brief passes\n- the handover is updated\n- one paragraph: what changed and why\n- nothing left half-finished on the branch",
      "markdown": "/templates/done.md",
      "page": "/#copy"
    },
    {
      "id": "stop",
      "name": "Stop rules",
      "when": "Anything hands-off",
      "summary": "Knowing when to stop matters more than knowing how to continue.",
      "format": "text/plain",
      "text": "Stop and ask if:\n- the same fix has failed twice\n- the task has grown past its brief\n- you'd need a new dependency or permission\n- you're about to guess at something a human knows",
      "markdown": "/templates/stop.md",
      "page": "/#copy"
    }
  ]
}
