create-better-spritesheet
Skill1 starsChanged 45 days agoMITView on GitHub
Create, rebuild, diagnose, or review high-fidelity 2D character sprite-sheet production, with complete motion-plan approval, deterministic smooth-raster rendering, and replayable delivery evidence.
In this repository18 other files it gives its agents.
- AGENTS.mdAGENTS.md
- codebase-designSkill
- eli5Skill
- find-skillsSkill
- improve-codebase-architectureSkill
- planning-with-filesSkill
- refine-promptsSkill
- show-meSkill
- app-logo-generatorSkill
- decompose-work-atomicallySkill
- download-online-videoSkill
- download-video-transcriptsSkill
- video-to-spritesheetSkill
- handoff-next-goalSkill
- handoff-next-taskSkill
- human-reasoningSkill
- refine-promptsSkill
- solve-from-first-principlesSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 5 are distinctive, and 0 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.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Use this Skill for character identity, animation planning, sprite-source review, deterministic sheet production, delivery verification, or …
- Distinctive: line 14
Use the stateful seam for production and read-only verification:
- Distinctive: line 20
Use the evidence seam for deterministic diagnostics, sealing, and independent replay:
- Distinctive: line 28
Read the references required by the current route:
- Distinctive: line 52
Stop at the current checkpoint whenever input, approval, or eligibility is missing.
≈ “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.
Files
Create Better Spritesheet
Use this Skill for character identity, animation planning, sprite-source review, deterministic sheet production, delivery verification, or diagnosis. Infer the route from the requested outcome and live artifacts; do not ask the user to choose an internal mode.
The installed create/rebuild adapter is smooth-raster/v2. Pixel art and other raster paradigms have different authoring and sampling rules: review them without conversion, but return UNSUPPORTED_CAPABILITY for create/rebuild until a matching production adapter exists.
Public seams
Use the stateful seam for production and read-only verification:
<python> <skill-dir>/scripts/spritesheet_production.py advance|verifyUse the evidence seam for deterministic diagnostics, sealing, and independent replay:
<python> <skill-dir>/scripts/spritesheet_delivery.py diagnose|seal-delivery|verifyResolve <python> to Python 3.10+ with NumPy and Pillow. Treat spritesheet_pipeline.py prepare-canonical|build-package|verify-package as an internal and compatibility adapter, not the primary production workflow.
Read the references required by the current route:
- art-direction-contract.md for identity, visual consistency, paradigm boundaries, and native-size review.
- motion-plan-contract.md for topology, animation principles, position roles, direction continuity, and the complete pre-image plan.
- approval-protocol.md for approval subjects, pauses, invalidation, and recovery.
- smooth-raster-profile.md for Alpha, outline, scaling, optical correction, and visual quality gates.
- production-interface.md for intent, checkpoint, response, and typed failure behavior.
- pixel-evidence.md for package v5 lineage, aliases, receipts, diagnostics, and replay boundaries.
- production-delivery.md for the sealed v2 evidence closure and claim classifications.
- reference-search.md only when usable action evidence is missing and discovery is wanted.
- runtime-integration.md only when the user explicitly authorizes changes to a target runtime.
- compatibility.md when inspecting or migrating v1/v4 jobs, packages, commands, or deliveries.
Create and rebuild workflow
- Resolve the production profile, target cell contract, required canonical views, art-direction contract, actions, directions, and delivery scope. Reuse already resolved outline settings unless current evidence conflicts.
- Prepare and replay-admit one high-resolution canonical per required direction-camera view. Inspect high-resolution and native-size previews on white, dark, and checkerboard backgrounds. Ask for canonical approval only when the complete view set is eligible.
- Build one complete
motion-plan/v2covering every clip and every logical playback position. Present the full current plan and obtain explicit approval before accepting or producing any keyframe or in-between image. Any material plan revision invalidates image work and requires presentation and approval of the complete revised plan. - Accept the planned high-resolution keyframe sources through raw RGBA admission. Present and approve the complete keyframe set.
- Accept the planned in-between sources, if any, through the same admission. Represent holds and loop closure as aliases with no new image source. Present and approve the complete ordered sequence.
- Build
spritesheet-package/v5, replay deterministic rendering and assembly, compute diagnostics, and enforce configured mechanical thresholds before opening package review. - Present the exact identity, motion plan, package, diagnostics, native-size board, onion skin, and playback previews. Record one complete package decision with an observation for every required subject.
- Seal
spritesheet-production-delivery/v2and independently verify its file closure, approvals, raw-source admissions, package replay, and diagnostics recomputed from final sheet pixels.
Stop at the current checkpoint whenever input, approval, or eligibility is missing. Do not generate images while the motion-plan gate is unresolved. Do not patch target cells or a finished sheet; correct the earliest owning source or contract and rebuild descendants.
Diagnose and review
Keep the subject byte-for-byte read-only. Verify from the outer delivery inward when available; otherwise start from the package manifest. Classify each conclusion as MACHINE-VERIFIED, REVIEWED, DECLARED, SUPPLIED, or unresolved. Report the earliest owning defect, affected descendants, and whether correction was requested; diagnosis alone does not authorize modification.
Completion
Report package-ready only when a sealed delivery passes independent verification. A bare package may be pixel-verified but has no delivery state. Separate machine results, human visual acceptance, declared intent, supplied external evidence, and unresolved uncertainty. Do not stage, commit, publish, replace runtime assets, or create a pull request unless separately requested.
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.

