norahe0304-art/30x-video / cli/skill
30x-web-to-video
FlaggedUse when a user gives a brand URL and wants a premium launch, demo, or marketing video. Zero-manual workflow — Claude runs the V2 orchestrator itself (scrapes brand evidence, auto-installs missing tools, generates the first-cut Remotion project). User only sees the URL and the Studio preview.
rules/finish-gate.md
View on GitHub<!--
[INPUT]: Scene drafts, BrandTruth, ScreenshotProof, VideoProof, taste rules, Impeccable anti-pattern vocabulary
[OUTPUT]: A blocking finish protocol before render with explicit rejection language and pass/fail checks
[POS]: rules/ 的哲学层出口; 把 taste-skill 与 Impeccable 的思想变成 render 前的强制门禁
[PROTOCOL]: 变更时更新此头部,然后检查 AGENTS.md
-->
# Impeccable Finish Gate
This gate is mandatory after preview and before render. It borrows the language of Impeccable and the discipline of taste-skill, but it does not require either tool to be installed. The job is simple: catch default bad solutions before they ship.
## Anti-Attractor Pass
Before render, force a short anti-attractor pass. Write down at least three reflex defaults the model wanted to reach for, for example:
- purple-gradient SaaS aura
- Inter-everywhere monotony
- card-stack crutch
- centered headline over busy UI
- generic "clean startup" layout
If you cannot name the reflex defaults, you are probably still inside them.
## What Must Be Written Down
Before render, the builder must be able to state:
1. **Default bad solution:** what the model would have done by reflex
2. **Why it was rejected:** what made it generic, monocultural, or unfaithful
3. **Brand truth override:** which real evidence forced a better decision
4. **Chosen direction:** what replaced the bad default and why it fits the brand
If this cannot be said clearly, the video is not ready.
## Monoculture Checks
Fail the render if any of these dominate the final video:
- Purple-gradient SaaS atmosphere with no brand justification
- Inter-everywhere monotony when the brand clearly signals another type attitude
- Card-on-card stacks or template dashboard grids used because they were easy
- Layouts that could belong to any AI startup with only a logo swap
- Decorative effects that have no information role
- Two or more scenes reuse the same centered-title / subtitle / CTA skeleton with only the copy changed
- Two or more scenes rely on the same three-card feature grid as a default reveal
- The overall cut reads as "generic AI output" before it reads as this brand
## Enterprise Credibility Checks
Fail the render if:
- The product UI looks ornamental instead of real
- Motion feels playful when the product needs trust
- Act 4 proof is weaker than the decorative transition budget
- The video hides behind abstract visuals because the product evidence is thin
- The close sounds like generic AI copy instead of a brand-specific claim
- Dense UI scenes are covered with headline text instead of being framed cleanly
- Count-ups, charts, or metrics animate so late that the viewer cannot verify them before the cut
## Template Trap Checks
Fail the render if:
- Two adjacent acts feel like the same layout with a palette swap
- The edit depends on decorative spacing rather than information hierarchy
- A scene survives the `logo swap test` with almost no other changes
- The most memorable visual move is a transition, not a product or proof moment
## Visual Discipline Checks
Audit the final cut for:
- **Hierarchy:** one clear focal point per frame
- **Hierarchy by two:** the hero element should win by at least two of size, contrast, position, or weight
- **Contrast:** text and proof surfaces are readable without straining
- **Whitespace:** negative space feels intentional, not empty or starved; reject `whitespace starvation`
- **Density:** UI scenes hold enough evidence to be persuasive but not noisy
- **Motion dignity:** transitions support the narrative instead of stealing attention
- **Remove test:** if covering an element improves the frame, that element is decorative filler and should be cut
## Scene Framing Checks
Reject scenes where:
- A dense UI tries to carry headline, subtitle, badge, CTA, and proof at once
- The eye has to choose between multiple hero candidates
- The scene needs blur, glow, or grain to hide weak composition
Preferred correction:
- headline frame first
- product frame second
- proof frame third
Do not pile all three into one compromised frame unless the composition remains obvious at a squint.
## Anti-Pattern Vocabulary
Name the problem precisely when rejecting a scene:
- `template layout`
- `cardocalypse`
- `inter everywhere`
- `purple reflex`
- `fake enterprise gloss`
- `headline over UI clutter`
- `proof starvation`
- `monoculture motion`
- `reflex default`
- `template trap`
- `generic AI output`
- `monoculture collapse`
- `card stack crutch`
- `hierarchy failure`
- `whitespace starvation`
- `decorative filler`
Precise naming matters. Vague criticism leads to vague fixes.
## Pass Conditions
The render may proceed only when:
- At least three reflex defaults were written down and explicitly rejected
- Screenshot proof is present in the story
- Video proof is used when available, or its absence is explicitly compensated for with screenshot-first staging
- At least two scenes in the body of the video feel product-specific rather than layout-specific
- The archetype sharpened the design without overpowering the brand
- No two acts collapse into the same template posture
- Dense UI scenes remain readable without headline clutter
- The team can explain why this does **not** look like a generic AI-generated launch video
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| planning-and-task-breakdownaddyosmani/agent-skillsBreaks work into ordered tasks. Use when you have a spec or clear requirements and need to break work into implementable tasks. Use when a task feels too large to start, when you need to estimate scope, or when parallel work is possible. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| screenpipethedotmack/claude-memSearch screen and audio history recorded by Screenpipe and optionally save selected context into claude-mem. Use when the user asks to recall desktop activity, a meeting, or something seen in another app through Screenpipe. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |
| contact-widgetnexu-io/open-designSelf-contained floating chat widget with welcome screen, social links, meeting button, and message input. Single HTML file, zero dependencies. | Stars: 108k | Updated | Claude CodeCodex | Verified | Install |
| meeting-notesnexu-io/open-designMeeting notes page — title bar with attendees, agenda checklist, decisions block, action items table with owners + dates, and a "next meeting" footer. Use when the brief mentions "meeting notes", "minutes", "1:1 notes", "all-hands recap", or "会议纪要". | Stars: 108k | Updated | Claude CodeCodex | Verified | Install |