agentleFS
PicsArtPicsArt/gen-ai-skills/skills/motion-review/SKILL.md

motion-review

Skill7 starsChanged 20 days agoMITView on GitHub

Review an existing scene for motion that is confusing, mistimed, distracting, broken at a seam, or unnecessarily expensive. Use when asked to critique or improve a finished animation, audit a multi-beat scene, prioritize motion fixes, or decide whether a static element should move, and before reporting that your own render, character, rig or acceptance video works.

3 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

8 rules. About 0 are common, 5 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    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
  1. Distinctive:

    Read the brief, storyboard, reference, or editor notes when supplied.

    line 23
  2. Very specific:

    Run picsart_scene_validate on the document under review.

    line 39
  3. Very specific:

    Use picsart_scene_query_layout at the start, interior, landing, and hold of each suspect motion.

    line 42
  4. Very specific:

    Use picsart_scene_contact_sheet with explicit times for decisive frames, including an interior frame rather than endpoints alone.

    line 47
  5. Distinctive:

    Avoid universal timing verdicts.

    line 92
  6. Distinctive:

    After an edit, validate again, recheck its entire beat, and inspect the neighboring seams.

    line 96
  7. Distinctive:

    Keep a short ordered set of actionable changes.

    line 107
  8. Distinctive:

    When the work matters (a change others will build on, a result shared with others), have someone other than the author review it with fresh …

    line 143

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Review motion in an existing scene

Review the authored scene and the moving picture together. A valid document and a good still frame cannot establish that a beat reads correctly in sequence. Find the few changes that would most improve comprehension, then verify each change in the interval where it matters.

This recipe reviews an existing document. For a new beat, start with motion-intent (fetch with picsart_scene_recipe, name motion-intent). For scene syntax and worked constructions, use scene-animation-reference and motion-pattern-library, fetched the same way.

Establish what the scene is trying to say

Read the brief, storyboard, reference, or editor notes when supplied. Without one, describe the apparent message as an inference and avoid treating your preferred style as the intended one.

Make a short timeline map from the scene itself: composition duration and frame rate, layer windows, track clips and transitions, motion tracks, looks, text animations, and nested scene_ref content. Use picsart_scene_inspect for a large document; it shows authored values and addresses, not evaluated root-time positions. Identify each visible beat's leader, the change it should communicate, and the interval in which that change must be read. Mark every seam and every long hold for closer inspection.

Do not add motion simply because a layer is static. Stillness may support reading, contrast, or an intentional reset. A proposed addition needs a communication job and a viewer-visible change it would clarify. Reject it when it only repeats another layer's gesture or competes with the leader.

Gather evidence at the right scale

  1. Run picsart_scene_validate on the document under review. Repair an invalid scene before judging subtle timing. A successful validation proves the document's contract, not the rendered beat.
  2. Use picsart_scene_query_layout at the start, interior, landing, and hold of each suspect motion. Query both sides of a seam. Map nested or layer-local clocks to root time before comparing values. If there is an explicit geometric expectation, use picsart_scene_check_contracts over its intended window, and picsart_scene_check_motion for frame-complete motion expectations.
  3. Use picsart_scene_contact_sheet with explicit times for decisive frames, including an interior frame rather than endpoints alone. Inspect pixels for visibility, crop, blur, text readability, and overlapping attention.
  4. Watch the continuous rendered interval at its intended viewing speed when the issue is pacing, acceleration, causality, or a seam. Stills can suggest where to look but cannot settle those questions. If only stills are available, mark the temporal judgment unverified.

For reference matching, compare the reference and candidate over the same window using picsart_media_track_motion_diff, picsart_media_flow_diff, or picsart_media_inventory_diff as appropriate. These answer different questions: trajectory, flow, or visible inventory. None judges whether an original motion choice is expressive.

Diagnose before changing a value

Name the observed failure in scene terms, then identify the smallest responsible span or layer.

ObservationFirst thing to inspectNext decision
The viewer cannot tell what changedleader, contrast, arrival order, holdremove competing motion or reveal the needed result more clearly
A move feels abrupt or drifts after landinginterior pose, easing, expression tail, layer clockcorrect its progress or settle at the intended root time
A button or pointer action feels disconnectedpointer arrival, response start, result startalign the causal event and inspect the full widget, not one child
The seam feels like a new slidecarrier identity, direction, overlapping windows, transition source frameschoose continuation, replacement, or an intentional reset before choosing a transition
Motion stutters or misses frame budgetactive video count, effect area, blur, live layer windowsreduce the heaviest live layers in that window and re-check the target's playback
A static interval seems deadreading demand, next action, intended pauseshorten the hold or add one purposeful change; a pause may be correct

These are hypotheses, not automatic fixes. Check the relevant recipe before changing a representation (fetch with picsart_scene_recipe): app-ui-animation for a simulated interaction, motion-scene-continuity for a seam, linked-element-motion for reflow. A preset cannot repair a wrong semantic relationship or an incorrect layer lifetime.

Choose the next fix

Rank findings by the viewer's likely misunderstanding, how much of the scene they affect, and the confidence of the evidence. Resolve structural failures and missing or incorrect content first, then causality and continuity, then timing and emphasis, then polish. When two changes could explain the same symptom, try the smaller one and re-render its window before adding more.

For each proposed change, record:

  • the layer or scoped target and root-time window;
  • the intended viewer-visible result;
  • the current observation and its evidence source;
  • one concrete edit to test, including removing motion when it has no job;
  • a numeric check where a numeric expectation exists, plus the interval to watch.

Avoid universal timing verdicts. A 0.2-second gesture can be a clear UI tap and an unreadable title entrance, depending on frame rate, travel, density, and context. Treat timing tables as starting trials. A rendered scene must be judged at its delivery frame rate and viewing size.

After an edit, validate again, recheck its entire beat, and inspect the neighboring seams. A successful midpoint frame does not prove the entrance, landing, or exit. When the deliverable has multiple output targets, verify the affected interval on each; report a target's translation diagnostic instead of assuming equivalent output.

Report a review

Lead with the highest-impact confirmed issue, or say that none was found. For each finding give a scene location, root-time interval, what the viewer sees, the evidence, and the proposed edit. Separate measured, visually inspected, and not yet verified claims. List rejected motion additions when the user asked where to animate, so stillness is an explicit design choice. Keep a short ordered set of actionable changes; do not manufacture polish findings to fill a list.

Accept your own render before reporting it

Apply this before you call any rendered scene, character, rig or acceptance video "working". The author of a render tends to confirm it; look for the ways it fails.

  1. Name the intended action in one sentence ("both hands hold the ball at chest height"). Judge the picture against that sentence, not against your previous draft or a metric.
  2. Lint the shot's window before you show it (picsart_scene_lint_layout; allow a pair meant to touch). The hosted lint reads layer geometry, not painted pixels: it warns on figures whose bounds overlap, on a figure the frame cuts, and on a layer standing farther back painted over a nearer one (groundPaintOrder), a character that floats small on a portrait frame (smallFigure, with the scale and floor-line bottom that stand it there), and a character drawn facing left placed with no scale, so it faces right (facingUnset: [-1, 1] faces it as drawn). It cannot see a prop its own body hides or a speaking mouth under its own arm or hand; look for those on the contact sheet yourself, along with a prop away from its hand, a figure or key layer the frame mostly misses, and frames that draw nothing but the background. A warning is a defect you found before the viewer: fix it, or allow it knowing why.
  3. Read each line's action against its words. For every spoken line, set the text beside the action keyed over it: a line that says "look" or "wave" plays that, not idle or talk. Nothing checks it.
  4. Look at the whole window. Make a contact sheet of at least 8 frames spread across the clip, including interior frames (picsart_scene_contact_sheet for a scene, picsart_media_video_stills for a video), and open it yourself. One still or the endpoints hide mid-motion tears, drift and pops. For sync between two things (mouth and voice, hand and prop), track both per frame (picsart_media_track_motion); stills hid a mouth drift that per-frame tracking caught.
  5. Check what the viewer sees, not what you built. A metric on the mechanism can pass while the picture is wrong: a grip socket landed within 0.55 px of the ball while the drawn hand stopped a hand-width short. Check the drawn result: contact between silhouettes, limbs bending the anatomical way, nothing passing through a body or prop, near and far parts in the right draw order, garments following the body without tears or flares. A mean pixel diff cannot see text or small parts.
  6. List defects first, each with the frame or time where it shows, then what works. One passing gate is not success. A first iteration may carry minor labelled defects but no bad mistake: a broken anatomy, a missed contact or a stated rule broken blocks it.
  7. Deliver the moving result. Send the labelled final video for motion, not a sheet of measurements; the sheet is your evidence, the video is the deliverable.

When the work matters (a change others will build on, a result shared with others), have someone other than the author review it with fresh eyes, applying this section.

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.