agentleFS
Sign inStart free
indranilbanerjeeindranilbanerjee/cutforge/skills/video-delivery/SKILL.md

video-delivery

Skill0 starsChanged 27 days agoMITView on GitHub

Deliver the video: encodes per destination, verified transfers and the project hand-off. \"export it for YouTube\"

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    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
  1. Very specific:

    Ask what the far side already has: before sending anything big.

    line 32
  2. Distinctive:

    Never delete or overwrite on the client's drives: Write into a new dedicated subfolder.

    line 41
  3. Very specific:

    Use references/screening-run-sheet.md.

    line 57

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Video delivery: the right file, verified on the far side, and a project anyone can continue

Delivery fails in boring ways: a 3.3 GB file on a FAT32 pen drive, a projector that cannot decode High@5.1, a transfer that dies at 80 %, a hand-off nobody else can rebuild. Plan for the slowest link and verify on the far side.

1. Encode per destination

python tools/deliver/encode.py <master.mp4> <preset> [out.mp4] (re-normalises loudness per preset):

PresetUse
projector1080p H.264 High@4.1 ~10 Mbps, -16 LUFS; small home projectors and TVs (bring projector720 as backup)
projector720720p High@3.1 ~6 Mbps; weak players, old TVs
youtube1080p 12 Mbps, closed GOP, -14 LUFS, faststart
reels1080×1920 30 fps, -14 LUFS (source must already be vertical; captions inside the safe zone)
whatsapp720p ~2.5 Mbps; send as a document to avoid recompression
broadcast1080p 20 Mbps, -23 LUFS (EBU R128); US: --lufs -24 --tp -2
archivethe master's video copied, audio 320k

Files over 4 GB do not fit FAT32: use exFAT/NTFS or play from a laptop (--fat32 warns). Platform specs and caption rules: ${CLAUDE_PLUGIN_ROOT}/docs/compliance-and-licensing.md section 2; projector notes: section 5.

2. Move big files safely

  • Ask what the far side already has before sending anything big. Real case: ten graded chapters (3.3 GB, 3.5 hours over the link) were cut losslessly out of the master already on the client's PC with tools/deliver/recover_chapters.py (ref on the source machine, cut on the far side), verified by decoded- frame md5; the mix went over as lossless FLAC (204 MB instead of 340 MB), sample-identical.
  • Split, send, join, verify: parts.py split <file> --mib 15 → send the parts → parts.py join <manifest.json> on the far side (resumable: run again until COMPLETE; it never deletes, it truncates to the last verified part and appends) → whole-file md5 → decode check (verify.py --final).
  • rcopy.py for resumable verified copies onto slow drives; xtar.py to extract archives on drives where delete/overwrite is not allowed, under a time budget.
  • Never delete or overwrite on the client's drives. Write into a new dedicated subfolder; list leftovers (transfer parts) for the client to delete themselves.

3. Hand over the project, not only the film

The final hand-off is a folder anyone (or any agent) can continue with one-chapter re-renders:

  • code (src/, tools/, film/, render.mjs, package.json), assets used (public/), narration (sound/vo/), cue sheet, graded chapters (out/ch/*_f.mp4), the mix, and the master
  • PROJECT.md, facts.md, decisions.md, sources.md, spend.md, all NOTES_vN.md
  • CLAUDE.md (pipeline commands and rules for any agent) + README.md (for a human) + config.json with the new machine's paths (folders, api_keys_file, gl, concurrency, browser_executable)
  • excluded: node_modules (npm install), models/ (fetch_models.py), raw chunks, transfer parts, keys
  • on the far side: npm install, python tools/setup_local.py, render one chapter's stills to prove it

4. The screening (personal films and events)

Use references/screening-run-sheet.md. In short: test on the exact device in the room days before; carry the master, a 720p backup and a phone copy; play from a laptop if the projector's player is unreliable; external speaker; 2 s of black at both ends; offline (no autoplay, no notifications); one person owns play/pause, the interval and the lights.

5. Archive

Master + project folder + sources.md (licences) + spend.md on two drives or a drive and a cloud account. Note the md5 of the master in PROJECT.md.

References

  • references/screening-run-sheet.md: device test, run sheet, backups, what can go wrong in the room
  • ${CLAUDE_PLUGIN_ROOT}/docs/lessons-learned.md section H

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.

Check your setup, free