video-creative-rules
Skill0 starsChanged 3 months agoMITView on GitHub
Creative rules and gates for a Video Studio cut. Read before planning a cut from a confirmed brief, before revealing text or syncing to music, and before preparing an export. Owns what good looks like (rubric, reading-time floor, beat and audio policy, delivery gates); it never decides direction, chooses a final or approves a preference.
In this repository23 other files it gives its agents.
- AGENTS.mdAGENTS.md
- domains/AGENTS.mdAGENTS.md
- index/AGENTS.mdAGENTS.md
- records/AGENTS.mdAGENTS.md
- research/AGENTS.mdAGENTS.md
- reviews/AGENTS.mdAGENTS.md
- skills/AGENTS.mdAGENTS.md
- ai-writing-reviewSkill
- general-writingSkill
- live-data-rewriteSkill
- live-ui-auditSkill
- parallel-build-sprintSkill
- r2cSkill
- research-stackSkill
- ship-itSkill
- skill-indexSkill
- streamdeck-authoringSkill
- wikipedia-researchSkill
- wikipedia-reviewSkill
- wikipedia-workflowSkill
- wikipedia-writingSkill
- writing-cleanupSkill
- writing-reviewSkill
The file
We split this skill into the rules it gives an agent
10 rules. About 0 are common, 7 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 22
If a rule below conflicts with the confirmed brief, the brief wins and the conflict is reported in the candidate summary.
- Distinctive: line 35
Use the source's own words and claims.
- Distinctive: line 42
Restraint over decoration: No stingers, risers, whooshes on every cut, lens flares, particles or visualizer graphics.
- Distinctive: line 44
Honest: No claim beyond the brief and the source.
- Very specific: line 56
Read references/plan-rubric.md.
- Distinctive: line 60
Commit to one angle and one hook.
- Distinctive: line 61
Write the storyboard with, per scene: what is on screen, every text line with its settled window, sequential or simulated interaction, …
- Distinctive: line 65
Write a rights line for every music, SFX and image asset.
- Very specific: line 79
Every reading-time or budget finding is either fixed or listed in the candidate summary with a reason.
- Very specific: line 85
Read references/delivery-gates.md.
≈ “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
Video creative rules
The confirmed brief owns direction. The human owns every creative decision. Studio owns revisions, exact export and review evidence. This skill owns the rules the production agent applies in between, and the checks that turn those rules into review findings.
Who decides what
| Layer | Owns | Never |
| Confirmed brief | audience, goal, constraints, required formats, workflow template | timing, copy, music |
| Production agent (this skill) | angle, storyboard, copy, timing, music and SFX selection, cue locks, poster frame | confirming a brief, choosing a final, approving a preference, publishing |
| Studio | revision ledger, exact export, sampled visual and audio review, repair scope | creative judgment |
| Human | every creative decision, the final, reusable preferences | nothing is inferred from silence |
If a rule below conflicts with the confirmed brief, the brief wins and the conflict is reported in the candidate summary. If a rule conflicts with a human instruction, the instruction wins.
The laws
- Short. The cut is as long as the brief's deliverable needs and not one scene longer. A launch or announcement is 15 to 25 seconds. A lesson segment is as long as its one idea. Every scene must earn its place.
- Readable. Pace comes from motion, cuts and light, never from pulling text before it can be read. Every line holds for its floor (references/reading-time.md). Fast in, then hold. Never fast in, then gone.
- Specific. The cut must feel made for this exact product, person or lesson. Use the source's own words and claims. No copy that could belong to any video.
- Show the real thing. At least one scene carries real footage, real UI or real copy from the source. Real frames beat recreations: capture the running product first, recreate only chrome. Abstract filler, stock motion and color washes are not scenes.
- The hook is the first two seconds. Plan the hook before anything else. It is an image, a line or a motion, and it decides whether the rest gets watched.
- Restraint over decoration. No stingers, risers, whooshes on every cut, lens flares, particles or visualizer graphics. Sound and motion support the story.
- Honest. No claim beyond the brief and the source. Keep the source's own caveats when it has them. Every music, SFX and image in the plan carries a source and a public-distribution status.
Shape to start from, not a template: hook (2 to 3s), reveal (2 to 4s), two or three highlights (5 to 12s), landing (2 to 4s). Adapt it to the brief's workflow (references/tone-presets.md).
The three gates
1. Plan gate (after confirmBrief, before the first edit)
Read references/plan-rubric.md. Then:
- Answer the rubric, including question 9 (entry, key action, result). If the source has no flow, write "no flow" and say what carries the cut instead.
- Commit to one angle and one hook.
- Write the storyboard with, per scene: what is on screen, every text line with its settled window, sequential or simulated interaction, audio intent, transition.
- Compute each scene's text budget (references/reading-time.md) and cut copy until every scene fits.
- Write a rights line for every music, SFX and image asset.
- Save the storyboard as
storyboard.jsonin the work directory and register it withregister_artifact. Runscripts/check-reading-time.mjs storyboard.json. A blocking finding means the plan is not done. Studio runs the same check on that file when it prepares a review, so its findings reach the human either way.
2. Cut gate (before requesting an export)
- Re-run the reading-time check against the timings you actually built.
- Beat locks: at most three strong-cue locks, marked in the plan with their time (references/beat-policy.md). Sequential readable lines never land on consecutive beats.
- Audio mix inside the policy (references/audio-policy.md); the game or product's own audio is preferred when its rights line allows it.
- Every reading-time or budget finding is either fixed or listed in the candidate
summary with a reason. Findings are
category: brief, so they need the human, not an unattended repair.
3. Delivery gate (export and handoff)
Read references/delivery-gates.md. Loudness is normalized to the target, the poster is a settled frame baked as frame 0, each required format is planned as a reframe with its own text budget, and the share copy is one caption in the brief's voice.
Files
| File | Read it to |
| references/plan-rubric.md | answer the nine questions and write the storyboard |
| references/reading-time.md | apply the floor, the budget and the two failure modes |
| references/beat-policy.md | lock reveals to music without hurting reading |
| references/audio-policy.md | set bed and SFX levels, choose cues, keep reactive visuals honest |
| references/delivery-gates.md | normalize loudness, pick and bake the poster, derive formats, write share copy |
| references/tone-presets.md | pacing, type and transition vocabulary per brief workflow |
| scripts/check-reading-time.mjs | turn the storyboard into review findings |
| scripts/loudness-gate.mjs | two-pass loudness normalization that keeps the video stream |
The storyboard format the checker reads:
{
"scenes": [
{
"id": "s2",
"start": 3.0,
"end": 7.5,
"texts": [
{
"id": "s2-card",
"text": "Ada planned it. Tuck built it. Pip checked it.",
"kind": "line",
"settledStart": 4.05,
"settledEnd": 7.5
}
]
}
]
}settledStart is when the text is fully in; settledEnd is when it starts to
leave. kind is label (three words or fewer) or line; it is inferred when
omitted. optional: true marks text that is not required reading (a footnote, a
HUD label): it is still checked against its own floor, but it does not count
toward the scene budget or the sequential spacing rule.
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.

