review
Skill473 starsChanged 45 days agoMITView on GitHub
Run the mechanical checks and the by-eye checks, then report what you found with evidence instead of declaring success. Use before finishing any pixel-art task, and when the user asks "is this good" or "why does this look off".
192 other projects use a version of this.
Adopted by
Most recently taken up by
- microsoft/veronatheir own version, first seen Nov 8, 2026
- microsoft/FluidFrameworktheir own version, first seen Nov 8, 2026
- charliee1w/consolidation-memorytheir own version, first seen Nov 8, 2026
- jisuzzzz/haechitheir own version, first seen Nov 8, 2026
- apokamo/kajitheir own version, first seen Nov 8, 2026
- dirien/yet-another-agent-harnesstheir own version, first seen Nov 8, 2026
Used with
Projects with it also use close-spec, design-with-images and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
dzhngSkillclose-spec
When a spec under specs/<feature>/ is done shipping (a write-spec build, or any planned task that produced a spec), archive it to specs/done/ and rewrite it from a build-plan into a durable rationale — the why, the principles, the invariants — pointing back to the real code for the how. Use when implementation has landed and the plan no longer matches what shipped, or the user says a feature/spec is finished. Pairs with [write-spec](../write-spec/SKILL.md) (the plan this closes) and [review](../review/SKILL.md) (run before closing).
Also in 5 projects that use this
dzhngSkilldesign-with-images
Explore UI and visual design with image generation, then iterate the real implementation against the user-selected concept. Use when brainstorming visual directions, offering mockup variants, or implementing an approved generated design; do not stop at the mockups or the first code approximation.
Also in 5 projects that use this
LambenthanSkilldaily-arxiv
每日从 arXiv 拉取新论文,过滤相关性,auto-ingest 高优先级论文,检测 SOTA 更新
Also in 5 projects that use this
LambenthanSkillexp-eval
实验判决门:Review LLM 独立评判实验结果 → 4 种判决路径 → 自动更新 claims confidence、ideas status、graph edges
Also in 5 projects that use this
- +3 more from dzhng
- +5 more from Lambenthan
Other versions192 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 192
- garrytan/gstack★ 141ksame name
- ansible/ansible★ 71ksame name
- docling-project/docling★ 70ksame name
- traefik/traefik★ 65ksame name
- ClickHouse/ClickHouse★ 51ksame name
- temporalio/temporal★ 24ksame name
- tisfeng/Easydict★ 15ksame name
- juicedata/juicefs★ 15ksame name
- elie222/inbox-zero★ 13ksame name
- gitkraken/vscode-gitlens★ 10ksame name
182 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
3 rules. 0 are common, 1 is distinctive, and 2 are very specific to this project.
0
Common
Most of the 157 edited versions we read keep these.
1
Distinctive
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
- Very specific: line 15
Before you look at the brief again: look op="preview", and write one sentence on what the picture shows to someone who was never told, plus …
- Very specific: line 64
If the sprite was drawn from a reference (reference op="list" is not empty), look op="compare" on each key frame too.
- Distinctive: line 110
Run the same checks, then name the specific cause rather than listing possibilities.
“Common” is measured: kept by at least 6 in 10 of the 157 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Review a sprite before calling it done
"Done" is a claim. This is how you earn it.
Procedure
0. Cold read
Before you look at the brief again: look op="preview", and write one
sentence on what the picture shows to someone who was never told, plus the
three things the eye lands on first. If that sentence does not name the
subject, or misses a part the brief makes central, the sprite fails review
whatever validate says. Squint at the same preview for value: the subject's
lit side should be among the lightest values and its core shadow among the
darkest; everything in the bottom third reads as murk.
1. Mechanical checks
validateReports off-palette colours, semi-transparent pixels, isolated stray pixels, untagged animation frames, uniform timing and unsaved state — each with a location where one exists.
Fix every error. For each warning you do not fix, have a reason and say it.
For animation, tell validate what the cycle is supposed to be, and it checks
that instead of only the generic rules:
validate expect={ layerFrames: { "sword-fx": [[3,4]] }, mustNotOverlap: [["sword","head"]] }layerFrames flags a layer that is empty inside its frame range or has pixels
outside it (a slash effect leaking into the idle); mustNotOverlap flags any
frame where two layers' pixels collide (the blade crossing the face).
2. Look at it
look op="preview"Work down rules://07-review-checklist:
- Silhouette — recognisable as one flat shape?
- Value contrast — would it survive desaturation?
- Light — one direction, everywhere?
- Ramps — do shadows hue-shift, or are they just darker?
- Lines — even runs, no doubled pixels?
- Outline — one style, applied consistently?
- Banding — long parallel stripes of adjacent ramp steps?
- Detail placement — spent on the face and weapon, not the boots?
If the sprite was drawn from a reference (reference op="list" is not empty),
look op="compare" on each key frame too. A mismatch the PixelSpec lists as a
deviation is a decision; any other large one — a limb in a different place, a
silhouette that lost its defining shape — is a finding.
3. Look closely at anything that felt off
look op="ascii"This is where you find the pixel one row too low and the line run of 3 in a sequence of 2s. If step 2 left you with "something is wrong but I cannot say what", this is the step that names it.
4. For animation
look op="filmstrip"Volume consistent? Height consistent? Does it loop? Do contact poses hold longer? Is there anticipation before strong actions?
Then check the in-betweens one at a time against their neighbours:
look op="onion" frame=3Ghosts of frames 2 and 4 under frame 3 show whether the spacing eases or jumps, and whether a limb changed length on the way — both invisible in a filmstrip.
5. Report
State what you checked, what you found, and what you chose not to change:
Validated clean. Silhouette reads at 1×; value contrast holds desaturated. Two compromises: the plume colour snapped from
#e04a6ato#ff77a8(ΔE 14) because PICO-8 has nothing closer, and the far arm is 2px thick rather than 3 to keep the silhouette from fusing with the torso.
A user can act on a stated compromise. They cannot act on one you hid.
When the user asks "why does this look off"
Run the same checks, then name the specific cause rather than listing possibilities. The usual answers, in order of frequency: too many colours, flat shading, semi-transparent pixels, mushy silhouette, uniform animation timing.
Related
rules://07-review-checklist, and every other rule it points at;
rules://92-generated-art-tells; and the Review section of the subject file
for what was drawn (skill://studio, Subject rules).
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.

