Frame qa
My daily-driver Claude Code and Codex loadout
npx -y skills add kennykankush/skillpack --skill frame-qaAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Review generated storyboard frames or contact sheets for product-launch motion quality. Use when the user asks whether frames look professional, cohesive, AI-video-ready, faithful to screenshots, or should be regenerated. Outputs keep/regenerate decisions and targeted regeneration notes; does not manually edit frames.
SKILL.md
3.7 KB, as published. Nobody here has run it
Frame QA - Storyboard Quality Gate
This skill reviews storyboard frames after generation.
It acts like a motion-design art director and production QA pass: keep what works, reject what damages product truth, and write precise regeneration direction.
Activation
Use this skill for requests like:
- "go through all the frames"
- "make sure these look professional"
- "vet against what we built"
- "which frames should regenerate?"
- "this looks amateurish"
- "does this honor the screenshots?"
- "is this good for Higgsfield / Omni?"
Do not use this skill to generate new frames unless the user explicitly asks for regeneration and videos:storyboard + imagegen are also used.
Inputs
Inspect as available:
- individual generated frames
- contact sheet
- prompt logs
- source product screenshots
- accepted prior versions
- rejected examples
- style contract
- narrative beat list
- target video tool constraints
Use visual inspection for image files. Do not judge from filenames alone.
Review Axes
Score each frame against:
- Product truth: real UI language, real brand, correct feature premise.
- Cohesion: belongs to the same sequence as the other frames.
- Motion plausibility: looks like a still from an actual launch animation.
- Composition: hierarchy, spacing, crop, negative space.
- Text quality: no gibberish, fake paragraphs, bad labels, or overcrowding.
- Logo/brand fidelity: no invented marks or mutated brand assets.
- AI-video readiness: stable enough for interpolation without hallucinated resets.
- Professional finish: not amateur, not generic, not overly decorative.
Decision Labels
Use only these labels:
keep: frame is production-usablekeep with note: usable, but needs a handoff constraintregenerate: violates product truth or qualityoptional: nice but not needed for the sequencereplace narrative: the frame is okay visually but the beat is wrong
Output Format
Default:
Frame QA
Overall:
<short read on the sequence>
Frame 01 - keep
Why:
Motion note:
Frame 02 - regenerate
Issue:
Regeneration direction:
Frame 03 - keep with note
Why:
Handoff constraint:
For many frames, use a compact table:
| Frame | Decision | Issue | Regeneration / Handoff Note |
| --- | --- | --- | --- |
Regeneration Direction
Good regeneration direction names the exact failure and the desired fix:
Regenerate frame 06 with upright UI, one centered leaderboard module, cleaner gradient background, no extra decorative cards, no generated paragraphs, and stronger continuity with frames 05 and 07.
Bad:
Make it better and more professional.
AI-Video Readiness Checks
Flag frames that may look good as stills but fail in video:
- too much angled UI
- dense text that will mutate
- end frame radically different from start frame
- important logo too small
- background with many decorative elements
- multiple UI modules competing
- lack of clean hold frame for stitching
Recommend either a simpler frame or a clearer motion-handoff constraint.
What Not To Do
- Do not manually edit frames.
- Do not create local composites to "fix" bad generations.
- Do not accept mutated logos or gibberish because the overall vibe is nice.
- Do not regenerate everything if only one or two frames break cohesion.
- Do not make the QA so broad that the user cannot act on it.