Debrief Templates

{
  "externalId": "support_weekly_123",
  "title": "Support Weekly Brief",
  "summary": "Queue health and escalation risk for support leads.",
  "status": "ready",
  "audience": {
    "type": "groups",
    "groupSlugs": ["support-leads"]
  },
  "contract": {
    "version": "atlas-debrief-v1",
    "title": "Support Weekly Brief"
  }
}

{
  "version": "atlas-debrief-v1",
  "title": "Monday Morning Brief",
  "summary": "A calm Monday kickoff with priorities, schedule pressure, and the first unblocker.",
  "atlas": {
    "openingSpeech": "I have your Monday morning brief ready. Let me know when you want to begin.",
    "instructions": "Keep this calm, practical, and short. Pause between sections.",
    "delivery": "manual"
  },
  "canvas": {
    "title": "Monday Morning Brief",
    "deliveryMode": "replace",
    "pages": [
      {
        "title": "Week Start",
        "html": "<div><h2>This week starts here</h2><div><strong>Primary focus:</strong> ship dashboard polish</div><div><strong>Calendar pressure:</strong> 2 decision meetings</div><div><strong>Watch item:</strong> OpenClaw bridge reliability</div></div>"
      },
      {
        "title": "Priority Board",
        "html": "<div><h2>Today's priority board</h2><table><tr><th>Now</th><th>Blocked</th><th>Needs decision</th></tr><tr><td>Finish room UI polish</td><td>OpenClaw retries</td><td>Debrief resume UX</td></tr></table></div>"
      },
      {
        "title": "Week Shape",
        "html": "<div><h2>Week shape</h2><p>Heavy decision load early. Protect build time on Wednesday and Thursday.</p></div>"
      }
    ]
  },
  "sections": [
    {
      "title": "Week Start",
      "speech": "This week has one clear priority, one operating risk, and two meetings that actually need preparation. The point is to start deliberate, not reactive.",
      "checkpointPrompt": "Let me know when you want today's priority board.",
      "pageIndex": 0,
      "delivery": "wait_for_prompt"
    },
    {
      "title": "Priority Board",
      "speech": "For today, keep the focus narrow: finish the visible dashboard polish, keep an eye on bridge reliability, and clear the one decision that might stall the rest of the week.",
      "checkpointPrompt": "I can show you the week shape next.",
      "pageIndex": 1,
      "delivery": "wait_for_prompt"
    },
    {
      "title": "Week Shape",
      "speech": "The week is front-loaded with meetings, so the main move is to protect execution time once the early decisions are out of the way.",
      "checkpointPrompt": "That is the Monday brief. If you want, I can revisit a page or turn this into an action list.",
      "pageIndex": 2,
      "delivery": "wait_for_prompt"
    }
  ],
  "actions": {
    "speakToRoom": true,
    "openCanvas": true
  }
}

{
  "version": "atlas-debrief-v1",
  "title": "Research Summary Debrief",
  "summary": "A research walk-through with the answer, strongest evidence, and the tradeoffs worth debating.",
  "atlas": {
    "openingSpeech": "I have the research summary ready. I can walk you through the answer, evidence, and recommendation when you want to start.",
    "instructions": "Answer first. Keep this crisp and evidence-led. Pause after each section.",
    "delivery": "manual"
  },
  "canvas": {
    "title": "Research Summary",
    "deliveryMode": "replace",
    "pages": [
      {
        "title": "Answer First",
        "html": "<div><h2>Recommendation</h2><p>The lightweight bridge model is the fastest path to local private tools without exposing localhost to the internet.</p></div>"
      },
      {
        "title": "Evidence",
        "html": "<div><h2>Evidence</h2><table><tr><th>Option</th><th>Setup friction</th><th>Security posture</th></tr><tr><td>Polling bridge</td><td>Low</td><td>Strong</td></tr><tr><td>Public tunnel</td><td>Medium</td><td>Weaker</td></tr><tr><td>Persistent socket</td><td>Medium</td><td>Strong</td></tr></table></div>"
      },
      {
        "title": "Tradeoffs",
        "html": "<div><h2>Tradeoffs</h2><p>Polling is simple and safe, but not instant. A persistent socket is smoother later, but more operationally complex now.</p></div>"
      },
      {
        "title": "Recommendation",
        "html": "<div><h2>Next move</h2><p>Ship the polling bridge first, then upgrade to long-poll or websocket only if latency or chatter becomes a real product issue.</p></div>"
      }
    ]
  },
  "sections": [
    {
      "title": "Answer First",
      "speech": "The short answer is yes: the local bridge model is the right choice. It keeps the setup simple, avoids public exposure, and fits the product well.",
      "checkpointPrompt": "I can show you the evidence next.",
      "pageIndex": 0,
      "delivery": "wait_for_prompt"
    },
    {
      "title": "Evidence",
      "speech": "The strongest evidence is the tradeoff profile. The polling bridge has the lowest setup friction while still keeping the security model clean.",
      "checkpointPrompt": "Next I’ll show you the tradeoffs.",
      "pageIndex": 1,
      "delivery": "wait_for_prompt"
    },
    {
      "title": "Tradeoffs",
      "speech": "The tradeoff is mostly experience versus simplicity. Polling is slightly less elegant than a socket, but it is much easier to ship and support first.",
      "checkpointPrompt": "I can finish with the recommendation.",
      "pageIndex": 2,
      "delivery": "wait_for_prompt"
    },
    {
      "title": "Recommendation",
      "speech": "The recommendation is to keep the current bridge, harden the product feedback, and only move to a more complex transport if the real usage pattern demands it.",
      "checkpointPrompt": "That is the research summary. If you want, I can turn it into a decision memo or next steps.",
      "pageIndex": 3,
      "delivery": "wait_for_prompt"
    }
  ],
  "actions": {
    "speakToRoom": true,
    "openCanvas": true
  }
}

{
  "version": "atlas-debrief-v1",
  "title": "Eval Debrief",
  "summary": "An eval review with score distribution, failure clusters, and the fixes that matter next.",
  "atlas": {
    "openingSpeech": "I have the eval debrief ready. Let me know when you want the overview.",
    "instructions": "Keep this direct and operational. Pause after each section.",
    "delivery": "manual"
  },
  "canvas": {
    "title": "Eval Debrief",
    "deliveryMode": "replace",
    "pages": [
      {
        "title": "Overview",
        "html": "<div><h2>Latest eval overview</h2><div><strong>Run status:</strong> completed</div><div><strong>Pass rate:</strong> 84%</div><div><strong>Failed cases:</strong> 6</div></div>"
      },
      {
        "title": "Score Distribution",
        "html": "<div><h2>Score distribution</h2><table><tr><th>Band</th><th>Count</th></tr><tr><td>90 to 100</td><td>18</td></tr><tr><td>80 to 89</td><td>16</td></tr><tr><td>Below 80</td><td>6</td></tr></table></div>"
      },
      {
        "title": "Failure Clusters",
        "html": "<div><h2>Failure clusters</h2><table><tr><th>Area</th><th>Why it matters</th></tr><tr><td>Citations</td><td>Weak evidence carry-through</td></tr><tr><td>Accuracy</td><td>Borderline answers still slip through</td></tr><tr><td>Follow-up handling</td><td>Context drops on longer threads</td></tr></table></div>"
      },
      {
        "title": "Recommended Fixes",
        "html": "<div><h2>Recommended fixes</h2><ol><li>Tighten retrieval prompts</li><li>Add more citation-focused eval cases</li><li>Rerun after context-handling fix</li></ol></div>"
      }
    ]
  },
  "sections": [
    {
      "title": "Overview",
      "speech": "The topline is solid but not finished. Pass rate is up, but there are still a small number of recurring misses that are worth fixing before calling this stable.",
      "checkpointPrompt": "Let me know when you want the score distribution.",
      "pageIndex": 0,
      "delivery": "wait_for_prompt"
    },
    {
      "title": "Score Distribution",
      "speech": "The distribution says most cases are healthy, but the tail still matters. The low-score band is small enough to inspect manually and big enough to justify another iteration.",
      "checkpointPrompt": "Next I’ll show you the failure clusters.",
      "pageIndex": 1,
      "delivery": "wait_for_prompt"
    },
    {
      "title": "Failure Clusters",
      "speech": "The misses are not random. They cluster around citations, borderline accuracy, and follow-up handling, which makes the next fixes fairly clear.",
      "checkpointPrompt": "I can finish with the recommended fixes.",
      "pageIndex": 2,
      "delivery": "wait_for_prompt"
    },
    {
      "title": "Recommended Fixes",
      "speech": "The next move is not a broad rewrite. Tighten retrieval, add citation-focused eval coverage, and rerun once the context fix lands.",
      "checkpointPrompt": "That is the eval debrief. If you want, I can revisit any page or turn this into a concrete follow-up plan.",
      "pageIndex": 3,
      "delivery": "wait_for_prompt"
    }
  ],
  "actions": {
    "speakToRoom": true,
    "openCanvas": true
  }
}