longform-to-content
Skill18 starsChanged 4 months agoMITView on GitHub
Use when a user provides a long-form recording (webinar, class, lecture, talk, podcast, meeting — 통영상, 웨비나 녹화, 강의 영상, 풀영상) and wants it edited, repurposed, or published — a polished full-length edit, vertical shorts/reels, YouTube thumbnails, or scheduled posting to YouTube/Instagram. Also use for "영상 편집해서 올려줘", "쇼츠 만들어줘", "repurpose this recording", or turning one video into a content series.
- Reads credentials
1 other project uses a version of this, 1 of them edited.
Adopted by
Most recently taken up by
- Retired-Projects/ai-marketing-kit-publicedited, first seen Jul 30, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 1
- Retired-Projects/ai-marketing-kit-public★ 099% same
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository27 other files it gives its agents.
- skills/longform-to-content/llms.txtllms.txt
- skills/youtube-edit-kit/llms.txtllms.txt
- ad-videoSkill
- analytics-setupSkill
- b2b-cold-emailSkill
- brand-guideSkill
- carousel-generatorSkill
- content-repurposeSkill
- crm-connectSkill
- cyranoSkill
- dans-adviceSkill
- e-blast-newsletterSkill
- first-principles-coachSkill
- go-viral-or-dieSkill
- humanizerSkill
- image-genSkill
- kit-onboardingSkill
- lead-magnetSkill
- organic-socialSkill
- paid-adsSkill
- ppt-slide-generatorSkill
- print-designSkill
- publish-checklistSkill
- seo-geo-setupSkill
- thumbnail-makerSkill
- yc-office-hoursSkill
- youtube-edit-kitSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 26
G0 — Setup: Verify EVERY tool in references/SETUP.md with its verify command.
- Very specific: line 35
G2 — Edit: Follow references/PIPELINE.md exactly: transcribe once (word-level, cached) → measure layout geometry per recording → EDL …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Longform → Content
One raw recording in → a published content system out: relayouted full edit + burned captions + cold open, 4–8 vertical shorts, CTR thumbnails, and scheduled posts.
Every gate below is MANDATORY, in order. Skipping a gate or a self-test is a failure, not a shortcut. Violating the letter of these rules is violating their spirit. Each rule exists because an agent already broke it in production (see Red Flags).
Gates (run in order)
digraph gates {
"G0 Setup check" -> "G1 Brand check" -> "G2 Edit pipeline" -> "G3 Thumbnails" -> "G4 Publish";
"G0 Setup check" -> "STOP: give user references/SETUP.md installs, wait" [label="tool missing"];
"G1 Brand check" -> "Run brand intake (references/BRAND.md)" [label="no DESIGN.md"];
"G4 Publish" -> "STOP: user has no Zernio → onboarding in references/PUBLISHING.md" [label="no account"];
}G0 — Setup. Verify EVERY tool in references/SETUP.md with its verify command. Any missing → do not improvise; hand the user that file's install links + instructions and wait. Never substitute a "similar" tool.
G1 — Brand. A DESIGN.md with measurable tokens (2 colors as hex, fonts as file
paths, wordmark rule) MUST exist before any pixel is rendered. None? Run the intake in
references/BRAND.md — derive from the user's site/deck, or interview, or apply the
neutral default. Never invent brand values silently.
G2 — Edit. Follow references/PIPELINE.md exactly: transcribe once (word-level, cached) → measure layout geometry per recording → EDL (silence + tight-pacing + disfluency/stumble cuts — tight is the retention default, PIPELINE.md §3–4) → relayout render → cold open + cards → corrected burned captions LAST → SRT + chapters → frame-by-frame self-review before showing the user. Caption timing/text correctness has its own bible: references/CAPTIONS.md — read it before writing any cue.
G3 — Thumbnails. references/THUMBNAILS.md. The main thumbnail's claim MUST be the first spoken line of the cold open, and the user's real face photo MUST be used.
G4 — Publish. Packaging (titles, descriptions, chapters, cadence) per references/YOUTUBE.md; mechanics per references/PUBLISHING.md via Zernio API. Publishing is user-gated: get explicit per-destination approval ("올려줘/schedule it") before any publishNow or scheduledFor call. On any timeout: LIST posts before retrying — never blind-retry.
Quick reference
| Task | Where |
| Tool installs + verify commands | references/SETUP.md |
| Brand intake / DESIGN.md template | references/BRAND.md |
| Full pipeline: cuts, disfluency, relayout, cold open | references/PIPELINE.md |
| Caption correctness: timing math, ASR corrections, styles | references/CAPTIONS.md |
| Shorts: moment selection, canvas zones, hook titles, first-frame rule | references/SHORTS.md |
| Thumbnail CTR rules + generation | references/THUMBNAILS.md |
| YouTube packaging: titles, descriptions, chapters, cadence, analytics | references/YOUTUBE.md |
| Zernio onboarding + API + scheduling | references/PUBLISHING.md |
| Reference implementations (adapt constants) | scripts/ + scripts/README.md |
| Agents without a skills harness (paste-in prompt + capability matrix) | BOOTSTRAP.md |
Secrets rule (absolute): skills, handoffs, and scripts reference secret NAMES
(ZERNIO_API_KEY) and where the user stores them — never values, never the user's
account IDs, never their key-file contents. Scrub before sharing anything portable.
Red Flags — STOP, you are about to repeat a real production failure
| Rationalization | Reality (it already happened) |
| "I'll put the absolute path in the ffmpeg filter" | C: colon breaks filter parsing. Relative path + cwd only. |
| "Python print is fine without utf-8" | cp1252 crashed on Korean. PYTHONUTF8=1 + reconfigure, always. |
| "The POST timed out, I'll just retry" | The post WAS created. Blind retry = duplicate publish. LIST first. |
| "ASR captions are probably fine" | Scribe mangled ~95 terms in one hour of Korean. Build corrections, grep-verify zero residuals. |
| "The image model will render the text correctly" | It dropped "$20" → "0". READ the output, regenerate on any glyph error. Never inpaint. |
| "Thumbnail and intro don't need to match" | Click → first line must repay the thumbnail claim. Reorder the cold open. |
| "I can skip frame review, the commands succeeded" | Silent failures are invisible in exit codes. Extract frames and READ them. |
| "One-pass filtergraph is simpler" | Double-encodes when overlays change. Per-segment extract → -c copy concat. |
| "Captions can go before the overlay" | Overlays hide captions. Captions are LAST. |
| "User has no brand, I'll just pick something" | Run the intake. Defaults are chosen BY RULE, not by taste. |
Scope limits
- Talking-head/presentation recordings. Not for music videos or animation-first work.
- Voice processing (denoise/EQ) is opt-in only — default keeps original audio untouched.
- The agent's ceiling is staged/scheduled posts; the user presses the final go.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

