{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreiad6pyxdj53k7vgy75zrcthegmgnbjcr345f6xxlgeoefizcmoktq",
    "uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mkl27tdy5h62"
  },
  "path": "/t/diep-flap-surgery-recovery-prompts-txdiepflap/1379877#post_4",
  "publishedAt": "2026-04-28T16:07:16.000Z",
  "site": "https://community.openai.com",
  "tags": [
    "@jerryshah"
  ],
  "textContent": "Welcome to the dev Community! @jerryshah\n\nThat’s a really valid concern, especially for something like post-surgical recovery where consistency actually matters.\n\nThis can happen because small changes in phrasing can shift how the model interprets intent, especially for complex domains like medical recovery.\n\nA few things that usually help improve consistency:\n\n  * Use a fixed structure in your prompt (e.g., “Always respond with sections: Overview, Timeline, Risks, Care Tips”)\n  * Set expectations explicitly (e.g., “Be detailed, structured, and avoid generalizations”)\n  * Provide a reference example of the exact format you want\n  * Keep key constraints repeated across prompts instead of relying on prior context\n  * Use system instructions to lock behavior (e.g., “You are a medical information assistant that provides structured recovery guidance in consistent sections”)\n\n\n\nFor sensitive topics like post-surgical care, it also helps to anchor the scope (e.g., “focus on general recovery patterns, not personalized medical advice”) so the model doesn’t shift tone or depth.\n\n~Smith",
  "title": "DIEP Flap Surgery Recovery Prompts | TXDIEPFLAP"
}