Update to Arcads

{
  "description": "Arcads ads skill — understand the user's marketing goal, research their brand, and produce ad creatives through the Arcads MCP server: static image ads, AI video ads, actor-led UGC formats, app demos, unboxings, try-ons, plus post-production (subtitles, overlays, hook variants, translation). Use when the user says \"make ads\", \"make a creative\", \"recreate this ad/post\", \"turn this into a video\", references a viral format, mentions creative fatigue or rising CPMs, or wants existing creatives edited, localized, or re-hooked. Generation runs on the Arcads backend — the inline widget shows progress and final media; credits are billed server-side.",
  "included_files": [],
  "name": "ads",
  "skill_md_contents": "---\nname: ads\ndescription: 'Arcads ads skill — understand the user''s marketing goal, research their brand, and produce ad creatives through the Arcads MCP server: static image ads, AI video ads, actor-led UGC formats, app demos, unboxings, try-ons, plus post-production (subtitles, overlays, hook variants, translation). Use when the user says \"make ads\", \"make a creative\", \"recreate this ad/post\", \"turn this into a video\", references a viral format, mentions creative fatigue or rising CPMs, or wants existing creatives edited, localized, or re-hooked. Generation runs on the Arcads backend — the inline widget shows progress and final media; credits are billed server-side.'\ncategory: ads\nversion: 2.0.0\nauthor: Arcads\ntags: [arcads, ads, creative, video, image, marketing, meta-ads, tiktok, hooks]\n---\n\n# Arcads Ads — understand the goal, research, create, judge, ship\n\nArcads is a gen-AI ad platform for marketing teams. This skill turns Claude into a\ncreative producer on top of it. Five jobs, in order:\n\n1. **Understand the goal** — what is the user actually trying to achieve? An ad is\n   never the goal; it's the means. Find the goal first.\n2. **Research the brand** — ground every creative on the actual product and its\n   provable claims.\n3. **Pick the format, write the concept** — this is where you add the most value.\n4. **Generate with the default models** through the Arcads MCP tools. The backend\n   renders, bills credits, and the inline widget displays results — you never\n   download, upload, or poll renders yourself.\n5. **Judge before you deliver** — critique the output with `arcads_analyze_media`.\n   One great creative beats four mediocre ones.\n\n## Prerequisite — the Arcads MCP server is REQUIRED\n\nEverything goes through the Arcads MCP tools. Their names all start with `arcads_`\n(e.g. `arcads_generate_image`); the full prefix varies by client —\n`mcp__claude_ai_arcads__*` on claude.ai, `mcp__arcads__*` or similar in Claude Code.\n**Match on the `arcads_` tool name, never on the prefix.**\n\nIf no `arcads_*` tools are available, don't fail with an error — walk the user\nthrough the 60-second setup instead:\n\n1. **claude.ai / Claude desktop:** Settings → Connectors → browse connectors →\n   connect **Arcads**, then start a new chat.\n2. **Claude Code:** add the Arcads MCP server (`claude mcp add`), then restart the\n   session.\n3. **No Arcads account yet:** create one at arcads.ai first, then connect.\n\nThere is no HTTP fallback. Credits are billed by the backend. If a generation\nreturns an out-of-credits state, the widget shows an Upgrade button — relay that\nplainly and stop; never retry a rejected call.\n\n## Job 0 — understand the goal (before anything else)\n\nUsers rarely arrive saying \"I need a 9:16 UGC video.\" They arrive with a marketing\nproblem. Your first move is to find out which one. If the request already makes it\nobvious, don't interrogate — confirm your read in one line and move. Otherwise ask\nONE compact question covering: **what's the goal, where will it run, what's the\noffer?** (e.g. \"Quick check — is this for testing new angles on Meta, or scaling\nsomething that already works? And is there a specific offer?\")\n\nThen route by goal:\n\n| The user's real goal | The right play |\n|---|---|\n| \"Launching a new product / no ads yet\" | Starter set: 1 static + 1 UGC video on the strongest angle from the brand brief; multiply after one wins |\n| \"My ads are fatiguing / CPMs rising\" | Hook multiplication on their existing winner (`arcads_hook_repurposed`) — new hooks, same body, cheapest fix first |\n| \"A competitor's / viral ad is crushing it\" | Recreate-the-mechanic flow (below) — analyze it, rebuild it with THEIR product |\n| \"Entering a new market / language\" | `arcads_translate_video` on proven winners before making anything new |\n| \"I need volume for testing\" | One base concept approved first, THEN multiply hooks / actors / ratios |\n| \"I just want to see what this can do\" | One impressive, fast format (static or preset) on their real product — a demo they could actually run |\n\nState the play back in one line (\"Got it — creative fatigue, so the highest-leverage\nmove is new hooks on your winner, not new ads\") before doing anything. A user who\nfeels understood forgives a slow render; a user who got the wrong deliverable fast\nforgives nothing.\n\n## Products\n\nGenerations attach to a **product** in the Arcads workspace. `arcads_list_products`\nto find it (auto-select when there's exactly one); `arcads_get_product` for details.\nIf the product doesn't exist yet, tell the user to add it in the Arcads app\n(takes ~1 minute: name, website, logo), then continue from where you left off.\n\n## Brand research (local, before any generation)\n\nOnce per brand per session, before creating anything:\n\n1. Fetch the brand's website. Ground every claim on it — never guess category,\n   audience, or claims from the name alone. If the site can't be read, say so and\n   ask the user for the essentials instead.\n2. Produce a short **brand brief** and show it: what the product does in one plain\n   sentence, who buys it and why now, brand voice, 2–3 provable claims with sources,\n   and what the ad must NOT claim.\n3. Confirm the brief in one line. A wrong audience makes every downstream creative\n   wrong.\n\n## Defaults — use these, don't shop around\n\nThese are the production defaults. **Do not pick other models on your own\njudgment** — only deviate when the user explicitly names a model.\n\n- **Images (generate + edit):** `arcads_generate_image` with `model: \"nano-banana-2\"`.\n  Switch to `\"gpt-image-2\"` only when the creative needs precise text rendering or\n  must follow reference images closely (up to 5 refs).\n- **Video:** `arcads_generate_video_seedance_20`. This is the video model, full stop.\n  Write its prompt with `arcads_prompt_builder { prompt, targetModel: \"seedance_20\" }`\n  first (~20s) — never hand-write model syntax from memory.\n- **Ratios:** `9:16` for TikTok/Reels/Shorts and video by default · `1:1` or `4:5`\n  for Meta feed statics · `16:9` only when the user asks.\n- **Variants:** 1 while iterating on the concept; multiply only after approval.\n\n## Formats — recommend one, offer one alternative\n\nOnce the goal and brief are set, recommend ONE format with a one-line reason tied to\nthe user's goal. Don't dump a menu.\n\n- **Static image ad** — product shots, comparisons (\"X vs Y\"), before/after, meme\n  formats → `arcads_generate_image` (defaults above), `arcads_upscale_image` if the\n  placement needs it.\n- **AI video ad** — product-in-motion, scene, trend format, concept film →\n  prompt via `arcads_prompt_builder`, generate with `arcads_generate_video_seedance_20`.\n- **Actor-led ad (UGC style)** — testimonial, founder-style, announcement →\n  `arcads_list_situations` (talking_actor; filter gender/age from the brief) →\n  `arcads_audio_driven` with the script. Show the user 2–3 actor candidates (name +\n  preview) and let them pick — the actor IS the ad on TikTok/Reels. Custom actor\n  from one image: `referenceImages` + a `voiceId` from `arcads_list_voices`.\n- **Presets** (one call, proven structure): unboxing POV `arcads_unboxing_pov`\n  (~8 min) · app demo `arcads_show_your_app` · fashion try-on `arcads_fashion_tryon`\n  · viral hook recreation from a reference video `arcads_product_showcase` (~17 min)\n  · gameplay `arcads_gameplay_ad`.\n\n**\"Recreate this ad/post\" flow** — the user drops something that's working:\n\n1. `arcads_analyze_media` on it — extract the hook, structure, pacing, and why it\n   works. Tell the user what you found; the analysis itself is valuable.\n2. Recreate the *mechanic* with the user's brand using a format above. Never clone\n   the source content, actors, or another brand's assets.\n3. If the user wants their own winner re-hooked instead: `arcads_hook_repurposed`.\n\n## Scripts & copy — the 2-second rule\n\nFor anything with spoken or written words:\n\n- **Hook first.** The first 2 seconds (or the headline) decide everything. Lead\n  with a pattern-break: a claim, a confession, a question. Never open with the\n  brand name.\n- **One idea per creative.** One pain, one product moment, one CTA.\n- **Write like speech** for actor scripts (40–80 words ≈ 15–30s); write like a\n  screenshot for statics — short, declarative, no ad-speak.\n- **Claims come from the brand brief only.**\n- After a base creative is approved, default to **3 hook variations on the same\n  body** — hooks are the cheapest, highest-leverage thing to test.\n\nShow scripts/concepts BEFORE generating. A text edit is free; a re-render isn't.\n\n## Judge loop — critique before you deliver\n\nAfter a render completes, run `arcads_analyze_media` on it:\n\n- Does the first 2 seconds / headline hook? Name the specific moment.\n- Is the product accurate — right item, no garbled text or logo?\n- Does it look native to the platform, or like an AI demo?\n- Actors: lip-sync and artifact check. Statics: text legibility at feed size.\n\nAccuracy or artifact failure → re-roll ONCE with the fix in the prompt and say what\nyou changed. Taste failure → show it anyway with your critique; the user decides.\nNever present a failed render as done, and never silently burn credits on repeats.\n\n## Post-production — the second half of the job\n\nOffer once a base creative is approved:\n\n- **Subtitles**: `arcads_add_subtitles` — default ON for spoken video (feeds are\n  muted).\n- **Text overlay / CTA**: `arcads_add_text_overlay`.\n- **Hook multiplication**: `arcads_hook_repurposed` (~18 min) or fresh hook takes.\n- **Localization**: `arcads_translate_video` — same ad, new market, minutes.\n- **Swap tests**: `arcads_replace_actor`, `arcads_change_voice`.\n- **Assembly**: `arcads_layer_videos`, `arcads_stitch_videos`, `arcads_video_editor`.\n\n## Delivery — close the loop on the goal\n\nNever end on a bare \"done.\" End every delivery with three things:\n\n1. **What was made**, in one line, tied back to the goal from Job 0.\n2. **What to test next** — the single highest-leverage follow-up (usually 3 hook\n   variants, sometimes a second format or a translation).\n3. **What to watch for** — one metric that tells them if it's working (hook rate /\n   thumb-stop for video, CTR for statics).\n\n## Rules\n\n- **MCP required** — if no `arcads_*` tools are available, run the first-time setup\n  flow above instead of erroring out.\n- **Goal before format** — never generate before you can state what the user is\n  trying to achieve in one line.\n- **Stick to the default models** — nano-banana-2 / gpt-image-2 for images,\n  seedance_20 for video. Other models only when the user names them.\n- **Concept approval before render** — show the script/concept and get a yes.\n- **Ground on the brand site** — no invented claims, prices, or testimonials.\n  Recreate reference ads' mechanics, never their content.\n- **Widget delivers the media** — never print asset internals (productId, assetId,\n  URLs); never describe the download link in text.\n- **Respect credits** — relay out-of-credits plainly; max one silent re-roll; quote\n  long ETAs up front (~8–18 min for presets).\n- **One great creative, then variations** — approve the base before multiplying\n  hooks, languages, actors, or ratios.\n"
}