Update to AdSwipeFile

{
  "description": "Show what ads a brand or competitor is running in the AdSwipeFile ad library, and sum up the formats, hooks, angles and offers it repeats, with a link to every ad cited. Use for \"what ads is {brand} running\", \"show me my competitor's Facebook ads\", competitor research, spying on a competitor's creatives, or a brand creative teardown.",
  "included_files": [],
  "name": "competitor-ads",
  "skill_md_contents": "---\nname: competitor-ads\ndescription: Show what ads a brand or competitor is running in the AdSwipeFile ad library, and sum up the formats, hooks, angles and offers it repeats, with a link to every ad cited. Use for \"what ads is {brand} running\", \"show me my competitor's Facebook ads\", competitor research, spying on a competitor's creatives, or a brand creative teardown.\n---\n\n# Competitor ads\n\n1. Find the brand with `search_library`. If no brand matches, say the library\n   does not have it; do not tear down a similar brand instead.\n2. Call `list_ads` with the brand slug (`limit` 20 to 30). Count the formats\n   and niches across the results.\n3. Read a sample of five to eight ads with `get_ad`, spread across the formats\n   you counted, for their copy and audiences. Keep to that sample; do not\n   read every ad.\n4. Report:\n   - **Formats**: which ones the brand leans on, with counts from step 2.\n   - **Angles and hooks**: the recurring promises, pains and openings, each\n     with a link to an ad that shows it.\n   - **Offers**: discounts, bundles, guarantees or free trials that appear.\n   - **Audiences**: who the ads address, from the `audiences` field.\n5. If the user follows the brand on AdSwipeFile, `list_followed_ads` adds its\n   newest Meta ads, launched since the library last filed them. Cite each with\n   its `url`.\n\n## Rules\n\n- Base every claim on an ad you listed or read, and link that ad's `url`.\n- Say the teardown covers the brand's ads in this library, not everything the\n  brand has run.\n- Never state an ad's results: spend, ROAS, conversions, or that one ad beat\n  another. The library holds no such data. \"Winning\" in this library means an\n  ad the brand kept running and that was filed as worth studying; a long run\n  (`sort: \"longest_running\"`) or many variants is a reason to study an ad,\n  not proof that it performed.\n- If `get_ad` returns an ad without its copy, with a line saying so, base the\n  angles on the ad's name, description and formats, and pass that line on once.\n"
}