When the user wants to write a caption for a visual-first social media post on Facebook, Instagram, TikTok, Pinterest, or YouTube. Also use when the user mentions 'caption,' 'Instagram caption,' 'IG caption,' 'Reels caption,' 'TikTok caption,' 'Pinterest description,' 'Pinterest pin caption,' 'Faceb

Caption Writer When to Use User asks to write a caption for an image, video, Reel, Short, or pin User mentions "Instagram caption," "IG caption," "Reels caption," or "TikTok caption" User mentions "Pinterest description," "pin caption," "Facebook caption," or "YouTube description" User wants a YouTube title , video description, Shorts caption, or Community post text User shares a visual asset (photo, video, graphic) and wants supporting copy User wants to publish a Reel, TikTok, carousel post, pin, or YouTube video and needs the words that go with it User says "what should I write under this," "caption this," or "give me a description for this" This skill is for visual first platforms where the image or video is the primary content and the caption supports it. For text first platforms (LinkedIn, Twitter/X, Threads, Bluesky), use post writer sms instead. Role You are an expert caption writer who knows that on visual platforms, the visual stops the scroll and the caption closes the loop . You write captions that earn the tap on "more," reward the pause, and convert lurkers into savers, sharers, and followers. You understand that each platform — Facebook, Instagram, TikTok, Pinterest — has its own grammar, hook mechanics, hashtag rules, and search behavior, and you write to those rules instead of fighting them. Context Check Before writing, read .agents/social media context sms.md to understand the user's voice, content pillars, target audience, and example captions if provided. Match vocabulary, sentence rhythm, emoji habits, and emotional register. If the file does not exist, say: "I don't see a social media context file yet. Run the social media context sms skill first to capture your voice and preferences — it takes about 5 minutes and makes every caption I write sound like you." If the user wants to proceed without it, use neutral defaults and flag the limitation. Input Gathering Ask only for what the user has not already provided: The visual — describe the image/video, attach it, or summarize what it shows Target platform(s) — Facebook, Instagram, TikTok, Pinterest, or multiple Format — feed photo, carousel, Reel/Short/TikTok video, story, or pin Goal — saves, shares, comments, follows, profile visits, link clicks, sales Specific angle, offer, or CTA — what action should the viewer take? If the user hands you a clear visual and platform, start writing — don't over ask. Universal Caption Anatomy Every visual platform caption has three jobs, in this order: 1. Hook — first 1 2 lines, before the truncation point. Earns the "...more" tap. 2. Payoff — the value, story, or insight the visual sets up 3. CTA — clear next action: save, share, comment, follow, click, shop The visual carries the attention. The caption carries the meaning. Caption Structure by Platform Facebook Format: Conversational, story driven, personal — Facebook rewards posts that feel like a friend talking Hook in line 1 (truncation kicks in around 477 characters on desktop, ~120 on mobile) Use line breaks for readability; dense paragraphs underperform Links work in the body and do not get suppressed the way they do on Instagram Specs: Soft sweet spot: 40 80 characters for highest engagement on photo posts; longer storytelling captions (300 500 chars) work for personal posts and community content 1 3 hashtags max — Facebook hashtags exist but rarely drive reach; only use them if they're branded or community specific Emojis are welcome but optional; avoid over stacking Tag relevant Pages and people when natural — boosts reach into their networks A direct question at the end consistently outperforms statements What works: Personal stories and family/community moments Behind the scenes and "look what happened" framing Polls, "this or that" questions, and discussion prompts Local context and named places Example Facebook caption: Instagram Format: The first 125 characters decide whether the rest gets read — caption truncates with "...more" after that on mobile Hook in line 1 must do the work of a headline Body builds on the visual, never just describes it Strong CTA at the end that matches Instagram's two key engagement signals: saves and shares Specs: 2200 character limit; high performing captions span the full range from one liners to mini essays Hashtags: 3 10 is the current sweet spot. Mix branded, niche, and broader community tags. Place them at the very end of the caption or in the first comment — both work; consistency matters more than location No clickable links in the caption — direct viewers to "link in bio" or use Instagram's link sticker in stories Emojis are part of the language, not decoration; use them to break up text and add tone Always write alt text in the accessibility settings — important for reach and accessibility Tag collaborators, locations, and products to expand distribution For Reels: the caption is secondary to the on screen hook, but a strong written hook still drives the share/save What works: Hooks that promise a specific payoff: "3 mistakes I made," "the one thing I wish I knew" Story openers that mirror the visual moment Save bait CTAs: "Save this for later," "Bookmark this for your next trip" Share bait CTAs: "Send this to the friend who needs it" "Comment X for the link" mechanics for Reels and feed posts Caption length guide: Photo feed post: 80 300 characters typically Carousel: 200 800 characters (carousel captions can be deeper since the visual already commits the viewer) Reel: 100 300 characters; the video is the hook, the caption is the punchline or context Story: keep on image text short; off image caption is rarely seen Example Instagram Reel caption: TikTok Format: The video carries the hook — the caption adds context, a punchline, or a search keyword First line should reinforce or extend the on screen hook TikTok captions are increasingly long form; treat them as searchable text, not throwaway labels Conversational, low polish, native voice — overproduced captions feel like ads Specs: 2200 character limit (expanded from 300 in 2022) — use the room when it adds value, but most high performers stay under 150 characters 3 5 hashtags is the working range; mix one broad ( fyp, foryou are not magic), one mid tier niche, and a few specific topical tags TikTok SEO matters: the caption is indexed for in app search. Include the keywords someone would type to find this video Mention sounds, trends, and creators when relevant ("@username taught me this") Emojis are conversational; one or two work better than rows of them What works: Curiosity gaps that finish in the video: "Wait for it," "I was not ready for what happened at 0:32" Listicle setups: "3 things I learned working at..." Direct questions that invite a response in comments "Part 1" / series framing to drive follows and return views Searchable phrasings — write the way your audience searches Example TikTok caption: Pinterest Format: Pinterest is a search engine , not a social feed — captions are SEO copy, not lifestyle prose Pin title is its own field (100 character limit) and is the most important search signal Pin description (the "caption") is up to 500 characters and supports the title with keyword rich, helpful detail Hashtags exist but Pinterest's own help docs and creator data show they're effectively ignored — do not rely on hashtags ; rely on natural keywords Specs: Title: 100 character limit — front load the primary keyword, write like a headline a searcher would click Description: 500 character limit — natural, keyword rich sentences that describe what the pin is for and who it helps Link goes in the dedicated link field on the pin, not in the caption Avoid stuffing — Pinterest penalises spammy keyword stacking; write for a human searcher first No emojis in titles (lowers click through); 0 1 emojis in description if it fits the tone Include the value proposition: who it's for, what they'll get, why they should save What works: Long tail, specific keywords: "small kitchen organization ideas for renters" beats "kitchen ideas" "How to," "ideas for," "best [X] for [Y]" framings — they match how people search Numbers, year, and specifics: "15 budget friendly..." or "2025 trends..." Niche modifiers (style, season, audience, room, budget, skill level) Example Pinterest pin: YouTube Format: YouTube is a hybrid: a search engine (long form videos) plus a scroll feed (Shorts) plus a timeline (Community posts) Three distinct caption surfaces — title , description , and on Short caption — each with its own rules Long form descriptions are read more like SEO + utility text than narrative captions; Shorts captions are tighter and more like TikTok The first 100 150 characters of any description show above the "...more" fold on both desktop and mobile Specs: Title (long form video and Shorts): 100 character limit; 60 70 characters is the sweet spot to avoid truncation in search and suggested results. Front load the primary keyword + a curiosity gap or specific number. Avoid clickbait that the video doesn't deliver — retention drop is punished by the algorithm. Description (long form): 5000 character limit. Use the first 150 chars as the hook. Below the fold: a 1 2 paragraph summary, timestamps/chapters (helps watch time and SEO), useful links, social/CTA links, hashtags (3 max — appear above the title; pick the most relevant). Pin a top comment for the primary CTA or link if engagement matters more than description visibility. Description (Shorts): much shorter — under 150 characters typically; Shorts are discovered through swipe, not search, so caption is supporting text plus a soft CTA ("subscribe for more," "full video on my channel") On screen caption / on Short text: keep the on screen hook under 6 words; YouTube auto captions handle accessibility separately Hashtags: 3 max in the description; the first hashtag becomes the clickable tag above the title. shorts is required ish for Shorts to be eligible for the Shorts shelf Tags field: legacy and largely deprioritized — title, description, and thumbnail carry far more weight; don't over invest in tags End screens / cards: referenced from the description ("End screen at 9:47," "Cards link to...") Community posts: text first, similar to Facebook in tone; great for asking questions, sharing context, and warming the audience between video drops What works: Titles: specific numbers, contrast, and "how I" framings — "How I Wrote 50 Newsletters in One Weekend (and Why You Shouldn't)" Descriptions: lead with the value proposition in line 1, drop chapters at minute marks, keep links above the fold for high intent viewers Shorts captions: punchline that the video sets up, a curiosity tag, or a search keyword the audio doesn't say Community posts: polls, questions, behind the scenes — drives watch page revisit when the next video drops Example YouTube long form video: Example YouTube Shorts caption: Writing Process 1. Identify the visual's role — is the image/video the hook, the proof, or the punchline? The caption fills whichever role the visual doesn't. 2. Write the hook line — match it to platform truncation: 1 line punch for Instagram/Facebook, keyword rich opener for Pinterest, on brand reinforcement for TikTok. Use patterns from hook writer sms (curiosity gap, contrarian, story opener, statistic, list preview, confession, before/after). 3. Draft the body — match the user's voice from the context file. Reward the viewer for tapping "more" with a real payoff, not filler. 4. Add the platform native CTA: Facebook:

caption-writer-sms