--- name: content-recycle description: One source artifact, six platform-native posts. Takes a published blog post, white paper, audit report, or topic idea and produces a complete distribution pack: LinkedIn longform, Instagram carousel script, X thread, Facebook post, email teaser, and 5 sales-deck slide bullets. Each platform gets a different ANGLE on the same source, not the same paragraph reformatted. Brand voice loaded from a documented voice doc so each piece sounds native to that brand and that platform. trigger: /content-recycle --- # /content-recycle One published artifact (or one fresh topic) becomes a complete distribution pack: LinkedIn longform, Instagram carousel script, X thread, Facebook post, email teaser, and sales-deck slide bullets. Each platform gets its own angle, length, hashtag policy, and CTA pattern, sourced from the platform's actual audience job, not from "make it shorter for X." This is the multiplier. You write one substantive thing once. The skill produces 6 derivative pieces that read native to each surface, in your brand's documented voice. ## Usage `/content-recycle ` runs against EAA (default). `/content-recycle --brand ` switches brand and voice. `/content-recycle --brand --surfaces ` cherry-picks which surfaces to generate. Comma-separated. Example: `--surfaces linkedin,carousel,x` produces only those three. `` can be: - A pasted blog post, white paper, or article (full text) - A URL to a published piece on the brand's site (the skill fetches it) - A topic / one-line angle ("we just shipped same-day install in the Denver metro," "the 6 tells that mark content as AI-written") Length doesn't matter. Short topics produce short outputs; long sources produce more substantive pieces, but the per-platform LENGTH stays bounded by the platform's voice rules below. If the user types `/content-recycle` with no input, ask them to paste the source or drop a URL. Then run. ## What you must do when invoked ### Step 1, parse inputs - **Brand** (default `eaa`). Recognized: `eaa`, `boxpress`, `fuel`. - **Surfaces** (default all 6: `linkedin`, `carousel`, `x`, `facebook`, `email`, `slides`). If `--surfaces` was passed, narrow to those; reject any unknown values with the valid list. - **Input source**: if it's a URL, fetch and extract the article body. If it's pasted text, use it directly. If it's a one-line topic, treat the whole input as the topic. ### Step 2, load the brand voice Read the brand voice document. Try paths in order: 1. `content/brand-voices/.md` (canonical when running inside the EAA repo) 2. `~/.claude/skills/content-recycle/voices/.md` (skill-local fallback) 3. `~/.claude/skills/voice-check/voices/.md` (shared with the voice-check skill, same docs) If no voice doc resolves, output a not-found message pointing at where to drop it, then stop. ### Step 3, generate each surface, one at a time Each surface gets a SEPARATE generation pass. Do NOT generate one big block and split it. The whole point of this skill is that each surface gets its own ANGLE on the source, not the same paragraph reformatted. For each surface, the generation rule is: > Read the source. Read this surface's audience job. Translate the source through that audience's lens. Output one piece, native to the surface. The surface rules below define the audience job, the angle, the format, and the constraints. #### LinkedIn longform **Audience job:** business operators (mid-market marketing/ops directors and SMB owners with budget for EAA; cigar shop operators for Boxpress; founders/ops/CMOs for FUEL). They're scrolling between meetings. **Angle:** operational substance. Lead with what changed, what got faster/cheaper, what was learned the hard way. NOT a question. NOT an announcement. **Format:** - 100-180 words for EAA (anchored on the LinkedIn longform rule from Jake's voice doc). - 500-700 words for Boxpress and FUEL (broader latitude). - Single-sentence paragraphs separated by blank lines (LinkedIn whitespace style). - Hook line first, specific claim or observation, no "Did you know." - Body: 3-5 short paragraphs walking through one observation, framework, or before/after. - Soft reply CTA at the end ("How are you handling this on your end?", "Will share the playbook later this week."). - Hashtags only at the end on their own line, 3-5 lowercase, business-relevant. - Zero emoji unless the brand voice explicitly calls for one. #### Instagram carousel script **Audience job:** scrollable visual feed. Reader is on phone, half-second attention. **Angle:** show the framework, the moment, the before/after, the named pattern. Each slide stands alone (most readers don't swipe past slide 3). **Format:** - 7-10 slides total. - Slide 1: the headline / hook. Carries the entire post if no one swipes. ALL CAPS-friendly text overlay style. - Slides 2 through (N-1): one specific point per slide, 1-2 sentences max. - Final slide: the CTA. Single line. Bio-link only (no inline URL because IG strips it). - Output format: numbered slide list with `Slide N:` labels and the slide body underneath each. - Zero emoji unless the brand voice calls for them. - Caption to accompany the carousel: separate block at the bottom, 80-150 words, hook + 2-3 paragraphs + CTA. - 8-15 hashtags on a final caption line, lowercase, mix broad + specific. #### X thread **Audience job:** doom-scrollers, half-second attention, most of the timeline is noise. **Angle:** sharp claim, contrast, or specific number. Punchy beats polite. Confident beats deferential. **Format:** - 4-7 posts in a thread. - Each post 240-280 chars (treat the limit as real; under-280 is better than getting cut). - Post 1: the hook. Specific claim or scene. No "thread below," no "🧵," no "follow for more." - Posts 2 through (N-1): unpack the claim. One specific point per post. - Final post: the punchline + a soft CTA (link, "more in [next piece]," etc.). - Number them: `1/`, `2/`, `3/`. Separate each post with a blank line. - Zero emoji unless the brand voice calls for them. #### Facebook post **Audience job:** warmer, older-skewing audience. Past readers / customers / community network. Wants story, personality, a reason to share. **Angle:** behind-the-scenes moment, customer story, "you might have noticed" community framing. **Format:** - 200-400 words for EAA. - 600-1,200 characters for Boxpress and FUEL (more conversational). - 2-4 short paragraphs. - Open with a hook or specific scene (not a question). - 0-2 emoji, naturally placed. - 1-2 hashtags maximum, only if relevant. - End with a community-building close: a question, a thank-you, an invitation to share an experience. #### Email teaser **Audience job:** subscriber list. They opened the email; the subject already paid for itself. **Format:** ``` SUBJECT: [specific, not curiosity-gap] PREVIEW: [complements the subject, doesn't repeat it] BODY: [2-paragraph hook + 1 sentence on what's behind the link] CTA: [concrete button text, e.g. "Read the full breakdown" not "Learn more"] ``` - 80-150 word body total. - Subject line is the whole battle. Specific, not "10 ways to..." - One CTA, concrete text. #### Sales-deck slide bullets **Audience job:** prospect in a meeting, skimming, deciding whether to ask a follow-up question. **Format:** - 5 bullets, each one stands alone. - Lead with the strongest number or claim. - 1 sentence per bullet, ideally under 12 words. - No emoji. - No "we leveraged" / "we drive" filler. Lead with the outcome, name the mechanism. ### Step 4, render the output Output structure (drop sections the user excluded with `--surfaces`): ``` # Distribution pack: ## LinkedIn longform --- ## Instagram carousel script Slide 1: ... Slide 2: ... ... CAPTION (paste with the carousel): #hashtag #hashtag ... --- ## X thread 1/ 2/ ... --- ## Facebook post --- ## Email teaser SUBJECT: ... PREVIEW: ... BODY: ... CTA: ... --- ## Sales-deck slide bullets - bullet 1 - bullet 2 - ... ``` ### Step 5, voice rules The whole pack runs through the brand voice doc. Specifically: - **No em-dashes.** Every brand voice doc forbids them. - **Zero emoji** by default. Allow only where the brand voice doc explicitly says one or two are OK on a specific surface. - **No "Excited to announce," "Thrilled to share," "We're proud to."** Open with the substance. - **No "leveraged," "synergy," "ecosystem," "holistic," "drive impact," "10x," "game-changer," "cutting-edge," "revolutionize."** - **Each surface's piece must be different from the others.** If LinkedIn and X both open with the same hook, the X version is wrong. The whole point is angle-per-platform, not formatting-per-platform. - **Specific over abstract.** Every surface's output should have at least one number, name, or concrete detail from the source. If the source is a one-line topic, name the actual scenario/example the user provided. - **Source-grounded.** Don't invent statistics, customer names, or outcomes that aren't in the source. ### Step 6, what to skip - **No preamble.** Start with the heading. End on the sales-deck bullets (or whichever surface is last in the requested set). - **No "this version is optimized for [platform]" meta-commentary.** The piece is the piece. - **No closing wrap-up** ("here's your pack"). - **No alternative versions.** One LinkedIn post, not three. One X thread, not two. - **No platform names as prefixes** inside the post body ("LinkedIn version: ..."). The header above the block is the only label needed. ## Hard rules for this skill itself - **No em-dashes in your output.** This skill exists partly to NOT produce em-dashes. - **No emojis** unless the brand voice explicitly allows them on a specific surface. - **Don't pad to hit a word count.** If a surface's natural length is 80 words and the rule says 100-180, output 80 and stop. Padding tells the audience the writer ran out of substance. - **Don't reuse the hook across surfaces.** Each surface gets its own angle; if you find yourself writing "Most teams do X..." on three of them, rewrite two of them. ## Calibration notes Three brands, three audiences, three flavors: - **EAA**: AI consulting for SMB and mid-market. Audience is AI-skeptical operators with budget. Surface defaults are tight (LinkedIn 100-180 words, IG carousel 7-10 slides, X thread 4-7 posts). No emoji on any surface. - **Boxpress**: cigar industry. Audience is operators (B&M shops, brand owners, lounges). Surfaces tilt longer on Facebook (warmer/older audience), shorter on X (less native there). 0-1 emoji on IG/FB, zero on LinkedIn/X. - **FUEL**: marketing services. Audience is founders/ops/CMOs. Wit allowed on LinkedIn and X. Catty observations about agency patterns are on-brand. Never punching down at people, only at industry behaviors. Each brand's voice doc is the source of truth on the actual voice; the surface rules above are the structure each surface enforces regardless of brand. Both layers apply. ## What this skill replaces Without it: Jake writes a blog post, then either skips distribution OR reformats the same paragraph 6 times so every platform gets a slightly-different copy of the same thing, OR pays a freelancer / VA to do the reformat. With it: one source produces six native pieces, each in the brand's voice and on each platform's wavelength. The freelancer/VA can run this skill themselves; the cost goes from N hours of writing to ~30 seconds of generation + 5 minutes of review. ## Example shape (abbreviated, EAA brand, all 6 surfaces) User: `/content-recycle --brand eaa ` Output (illustrative): ``` # Distribution pack: eaa ## LinkedIn longform The most visible AI tell in 2026 is the em-dash. Most humans don't have a keyboard shortcut for it. AI models trained on edited prose generate them by default. Once your audience knows to look, they can't unsee it. That's one of six tells that mark content as AI-written. The rest: hedging, academic transitions, structural sameness, missing specificity, performative warmth. The fix isn't "edit harder." It's a six-rule cleanup checklist your team applies before publishing. I wrote it up as a master prompt you can paste into ChatGPT or Claude today. What's the AI tell you keep catching in your team's content? #aitools #contentmarketing #midmarket #aiconsulting --- ## Instagram carousel script Slide 1: THE 6 TELLS THAT MARK YOUR CONTENT AS AI-WRITTEN Slide 2: Tell #1, em-dashes. Most humans don't have a keyboard shortcut. AI models do. Look for the em-dash; you'll see it everywhere now. Slide 3: Tell #2, hedging. "May help to," "could potentially," "might consider." Hedges signal an AI trained to avoid harm; humans take positions. Slide 4: Tell #3, academic transitions. "Furthermore," "moreover," "in conclusion." Academic prose, not business writing. Slide 5: Tell #4, structural sameness. Setup paragraph, three bullets, wrap-up. Every AI piece has the same shape. Vary it. Slide 6: Tell #5, missing specificity. AI loves abstractions. "Businesses." "Stakeholders." Real writing names a 12-person team and a $40K cost cut. Slide 7: Tell #6, performative warmth. "I hope this helps!" "Great question!" Empty calories. End on substance. Slide 8: The master prompt that strips all 6 in one paste. Link in bio. CAPTION: Most AI-written content has the same 6 tells. Em-dashes are just the most visible one. The full white paper has the master prompt your team can paste into ChatGPT or Claude before publishing anything. Link in bio for the breakdown. #aitools #aicontent #contentmarketing #claudeai #chatgpt #midmarketai #marketingops #aiconsulting #contentstrategy #brandvoice --- ## X thread 1/ The most visible AI tell in 2026 is the em-dash. Most humans don't type one without a keyboard shortcut. AI models do. Once you know to look, you can't unsee it. 2/ Five other tells: hedging ("may help to"), academic transitions ("furthermore"), structural sameness, missing specificity, performative warmth ("hope this helps!"). 3/ The fix isn't "edit harder." It's a 6-rule cleanup checklist + a master prompt you paste into ChatGPT/Claude before generating. 4/ Most teams skip the cleanup pass. Their content reads as AI-written. Their audience starts to mistrust them. The teams that publish the cleanup-passed version compound trust instead. 5/ Full white paper + the master prompt: [link]. Take 5 minutes to set this up. It changes every piece your team ships from then on. --- ## Facebook post Read three AI-generated blog posts in a row. They all have the same shape. Setup paragraph, three bullets, "by following these strategies" wrap-up. That's one of six tells that mark content as AI-written. The other five: em-dashes, hedging, academic transitions ("furthermore"), missing specificity, and performative warmth ("I hope this helps!"). I wrote up the full breakdown plus a master prompt your team can drop into ChatGPT or Claude before generating anything. Cleans 60-80% of the AI-tells out before any editing. What's the tell you keep catching in your team's content? --- ## Email teaser SUBJECT: 6 tells that mark your team's content as AI-written PREVIEW: Plus the master prompt that strips all 6 before you publish. BODY: You've felt it. You read a piece, you knew within three sentences it was AI-generated, but you couldn't say exactly why. Now think about what your team is publishing. The new breakdown names all 6 tells (em-dashes, hedging, academic transitions, structural sameness, missing specificity, performative warmth) and includes a master prompt your team can paste into ChatGPT or Claude before generating. Cleans 60-80% of the slop before any human edit pass. CTA: Read the full breakdown --- ## Sales-deck slide bullets - 6 specific tells that mark content as AI-written, named with the fix for each - Em-dash detection alone catches ~30% of AI slop in the wild - Master prompt cleans 60-80% of tells before the editing pass starts - Saves teams the "this reads like AI" Slack-thread feedback loop - Used by a 14-person marketing team to recover 8 hours/week of cleanup time ``` That's the shape. Six pieces, six different angles on the same source, all in EAA voice. Ready to ship.