ai-marketing-kit / longform-to-content

# longform-to-content — instructions for AI agents arriving via link

You are an AI agent and someone handed you this repository URL with a video to process.
This repo is a complete, mandatory-rules pipeline: one long-form recording in →
full edit + vertical shorts + thumbnails + scheduled publishing out.

DO THIS, IN ORDER:
1. Clone: `git clone https://github.com/crealwork/longform-to-content`
   (into your skills dir if you have one: ~/.claude/skills/ or ~/.codex/skills/)
   No shell? Read everything via the raw URLs below instead.
2. Read SKILL.md (gates G0→G4, red flags). Then read BOOTSTRAP.md and adopt its
   operating rules verbatim — including the vision-fallback rule (if you cannot view
   images, the user becomes your eyes; you never skip a visual self-test).
3. If the user has not given you the recording path yet, ask for it.
4. Start Gate 0: run every VERIFY command in references/SETUP.md, report a pass/fail
   table, and guide the user through installs for anything missing.
5. Do not render before the strategy confirmation (references/PIPELINE.md §2).
   Do not publish anything without explicit per-destination user approval.

FILE MAP (raw URLs, base: https://raw.githubusercontent.com/crealwork/longform-to-content/master/)
- SKILL.md                     — entry point: gates + red flags
- BOOTSTRAP.md                 — operating rules + agent capability matrix
- references/SETUP.md          — required tools, install links, verify commands
- references/BRAND.md          — brand intake when the user has no guideline → DESIGN.md
- references/PIPELINE.md       — the full edit pipeline (EDL, disfluency, relayout, cold open)
- references/CAPTIONS.md       — caption timing math, ASR correction dictionary, styles
- references/SHORTS.md         — shorts selection, canvas zones, hook titles, first-frame rule
- references/THUMBNAILS.md     — CTR rules, face-reference generation, thumbnail-intro match
- references/YOUTUBE.md        — titles, descriptions, chapters, cadence, analytics loop
- references/PUBLISHING.md     — Zernio onboarding, API, duplicate-publish safety
- scripts/README.md            — script map + placeholder legend
- scripts/*.py                 — reference implementations (measure_layout.py and
                                 scan_slides.py are ready-to-run; the rest need constants)

Every rule marked mandatory is mandatory. When a number in the skill conflicts with
your instinct, the number wins. When the user's explicit instruction conflicts with
the skill, the user wins — log the deviation.