본문으로 건너뛰기
cccskills
무료GitHub에 공개

sprite-forge

Generate game sprites, 8-direction turnarounds, walk-cycle sprite sheets, SVG characters, ASCII art and animated mascots from images or descriptions, with September 2026 providers: Retro Diffusion (pixel art, 8-dir rotation, frame animation, free palette tools), fal genmedia (GPT Image 2.5, Nano Banana 2, Kling 3, Seedance 2.5, Wan 3, Hailuo 03). Deterministic utilities: feet-line frame alignment, palette-preserving Scale2x/3x/4x pixel scaler, atlas manifest v2 with states/fps/loop/anchor, Aseprite JSON, pipeline profiles. Triggers on sprite, pixel art, sprite sheet, spritesheet, turnaround, 8-direction, walk cycle, atlas manifest, frame alignment, pixel scaler, Retro Diffusion, genmedia, SVG character, ASCII art, mascot, game asset, sprite pipeline profile.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

설치 방법 보기

포함된 파일(41)

  • SKILL.md9.5 KB
  • .gitignore25 B
  • assets/ascii-animation-template.tsx2.0 KB
  • evals/evals.json2.8 KB
  • pipelines/example.json2.9 KB
  • pipelines/schema.json4.0 KB
  • README.md5.5 KB
  • references/animated-ascii-sprites.md5.9 KB
  • references/ascii-animation-components.md5.1 KB
  • references/ascii-art-techniques.md2.7 KB
  • references/ayotomcs-deconstruction.md5.1 KB
  • references/chroma-key-transparency.md2.4 KB
  • references/frame-alignment.md3.6 KB
  • references/gemini-svg-generation.md3.9 KB
  • references/gsap-timeline-patterns.md2.8 KB
  • references/landscape-2026-09.md4.2 KB
  • references/lil-agents-character-spec.md3.5 KB
  • references/pipelines/open-rebellion.md1.4 KB
  • references/pixel-art-svgs.md2.2 KB
  • references/provider-matrix.md5.5 KB
  • references/turnaround.md3.2 KB
  • references/video-to-spritesheet.md4.5 KB
  • scripts/_images.py932 B
  • scripts/_secrets.py1.9 KB
  • scripts/align_frames.py13.2 KB
  • scripts/animation_builder.py13.1 KB
  • scripts/chroma_key.py3.3 KB
  • scripts/gen_image.py10.7 KB
  • scripts/generate_walk_video.py13.4 KB
  • scripts/image_to_ascii.py4.0 KB
  • scripts/pixel_art_generator.py9.8 KB
  • scripts/pixel_scale.py10.5 KB
  • scripts/retro_diffusion.py18.9 KB
  • scripts/run_pipeline.py13.7 KB
  • scripts/split_spritesheet.py2.7 KB
  • scripts/stitch_spritesheet.py14.4 KB
  • scripts/turnaround.py8.9 KB
  • scripts/video_to_spritesheet.py9.9 KB
  • templates/ascii-sprite-capture.html6.1 KB
  • templates/mascot-standalone.html1.1 KB
  • tests/test_deterministic.py6.0 KB

SKILL.md(원문)

설치하기 전에 에이전트에게 전달되는 지시 내용을 확인할 수 있습니다.

Sprite Forge

Game-ready sprites from images or text, September 2026 edition. Five pipelines, two wired providers (Retro Diffusion, fal genmedia), and a profile runner so a project can declare its own pipeline.

Cost rule: every provider script is a free dry run until you pass --yes. Retro Diffusion runs check_cost; genmedia prints the exact command and price. Never add --yes on a user's behalf without stating the estimate.

When to use

  • Pixel-art characters, 8-direction rotations, animated frames from a start image
  • Walk/idle/attack sprite sheets from video (first+last-frame loops)
  • HD painted turnarounds via GPT Image 2.5 edits
  • Upscaling indexed sprites without new colours (faithful-HD rule)
  • Atlas metadata with states, fps, loop, anchor (Phaser/Godot/Flame/Aseprite)
  • SVG mascots with GSAP, ASCII art, lil-agents dock videos (unchanged niche)

Not for: 3D generation (meshy skill), photoreal image editing (nano-banana-pro skill), generated sprites where original assets exist and parity is required (see references/pipelines/open-rebellion.md).

Work sequentially inside a pipeline; parallel agents only for independent characters.

Quick start

S=~/.claude/skills/sprite-forge/scripts
# 8-direction pixel sprite (one Retro Diffusion call, $0.25)
python3 $S/turnaround.py --mode pixel --prompt "astromech droid" --ref droid.png --out ./turn   # add --yes
# Walk cycle: Wan 3 clip on chroma -> key -> feet-align -> atlas v2 (+ Aseprite JSON)
python3 $S/video_to_spritesheet.py --provider genmedia --start-frame hero.png --end-frame hero.png \
  --prompt "side-view walk cycle in place, flat #00FF00 background" --duration 4 --cell-size 128x128 \
  --dedup 6 --state walk:12:loop --aseprite-json -o walk.png                                   # add --yes
# Indexed sprite 4x with the same palette
python3 $S/pixel_scale.py --input sprite.png --output sprite_4x.png --mode scale4x
# Frames -> sheet with manifest v2
python3 $S/stitch_spritesheet.py --input-dir frames/ --atlas v2 --state idle:8:loop --cols 8 -o idle.png
# Profile
python3 $S/run_pipeline.py ~/.claude/skills/sprite-forge/pipelines/example.json --dry-run

Pipeline 1: Turnaround (turnaround.py)

  • --mode pixel: rd_animation__8_dir_rotation (80x80, <= 5 refs) -> split -> stitch --atlas v2 --directions 8. One paid call.
  • --mode hd: anchor -> 4 cardinal edits -> 4 diagonal edits (anchor + two neighbours as refs) -> key if --chroma -> align_frames -> stitch. Default model openai/gpt-image-2.5/sunburst/edit (true alpha); --model fal-ai/nano-banana-2/edit --chroma for typed refs.
  • Order in the sheet: S, SE, E, NE, N, NW, W, SW. See references/turnaround.md.

Pipeline 2: Video -> sprite sheet (video_to_spritesheet.py)

[genmedia clip] -> ffmpeg -> chroma_key (once) -> align_frames (feet line, dedup, loop check) -> stitch v2. Default model alibaba/wan-3.0/image-to-video with --end-frame = start frame for a seamless loop; alternates Seedance 2.5, Kling 3 (@Element identity), Hailuo 03 (timecoded multi-state). Per-frame rembg on video is refused (matte flicker) unless --allow-rembg-video. Recipe and model table: references/video-to-spritesheet.md; jitter rules: references/frame-alignment.md.

Pipeline 3: Pixel scaling and palette (deterministic)

  • pixel_scale.py --mode nearest|scale2x|scale3x|scale4x, mode-P in, mode-P out, palette block and transparency kept, colour count never grows. --family dir/ --output-dir out/ transforms a frame family and writes family.json with a family sha256.
  • Retro Diffusion free tools for the generative side: retro_diffusion.py tool k_centroid_downscale --input-image in.png --field width=64 --field height=64, tool palette_converter --palette pal.png, tool color_reducer --field color_count=16, tool pixel_correction; pixelate (rd_pro__pixelate, $0.18) turns any raster into native pixel art.
  • pixel_art_generator.py survives only for the SVG <rect> mascot path (Pipeline 4).

Pipeline 4: SVG characters and GSAP mascots

Gemini SVG-as-code for clean vectors (references/gemini-svg-generation.md); pixel_art_generator.py --grid 16-24 --colors 4-6 --merge for rect-based pixel mascots (references/pixel-art-svgs.md); animation_builder.py --preset idle|bounce|lean|wave|walk|walk-and-bounce|walk-and-wave|typing for GSAP timelines (references/gsap-timeline-patterns.md, references/ayotomcs-deconstruction.md); generate_walk_video.py sheet.png --cols 6 --name robot for lil-agents HEVC-alpha dock videos (references/lil-agents-character-spec.md). Trim AI images first: magick in.png -fuzz 15% -trim +repage out.png.

Pipeline 5: ASCII

image_to_ascii.py img.png --mode gray|color|jp2a --width 80 for static art (references/ascii-art-techniques.md); frame-based React components and the asciistudio catalog in references/ascii-animation-components.md and references/animated-ascii-sprites.md; template assets/ascii-animation-template.tsx.

Profiles (pipelines/)

A profile (pipelines/schema.json) is an ordered list of stages over these scripts with ${var} substitution. run_pipeline.py profile.json [--dry-run] [--yes] [--stage name] [--from name] [--force] writes <output_root>/run/<id>/manifest.json with per-stage commands and input/output sha256 and skips unchanged stages. pipelines/example.json is the generic character profile; project profiles live in their repos (Open Rebellion: references/pipelines/open-rebellion.md).

Provider matrix (short)

NeedUseAlpha
Pixel still / identityRD rd_pro__default + --ref (<= 9), $0.18--remove-bg
8 directions, pixelRD rotate8, $0.25yes
Animate a start frameRD `animate --action walkingidle
HD anchor / editsopenai/gpt-image-2.5/sunburst[/edit] via gen_image.pyyes
Many typed refsfal-ai/nano-banana-2/editno, chroma
Loopsalibaba/wan-3.0/image-to-video first+lastno, chroma
Identity-critical clipfal-ai/kling-video/v3/pro/image-to-video @Element1no, chroma

Full table, prices, setup commands and PixelLab notes: references/provider-matrix.md. Setup: genmedia CLI (curl https://genmedia.sh/install -fsS | bash, then genmedia setup --non-interactive --api-key "$FAL_KEY"), RETRODIFFUSION_API_KEY in ~/.config/env/secrets.env; optional RD MCP claude mcp add --transport http retro-diffusion https://mcp.retrodiffusion.ai/mcp --header "Authorization: Bearer rdpk-...".

Gotchas

  • Feet line, not bbox. Halos and shadows move the bbox; align_frames.py scans for the lowest row with >= 5% solid width. Check feet_line_variance < 1.
  • One matte per clip. Chroma-key video frames with one setting; per-frame rembg flickers (perimeter_cv > 0.15).
  • RD limits. Styles cap at 384 px, animations at 256 and must match the start frame; input_image must be RGB without alpha (scripts composite onto --matte); negative is ignored; never write "pixel art" in RD prompts; balance is charged before generation.
  • genmedia. Use --async for anything over ~5 s (the video script does); verify endpoint ids and field names with genmedia models / genmedia schema; the GPT Image 2.5 endpoint spelling came from a third-party commit.
  • Nano Banana 2 still has no alpha; GPT Image 2.5 and RD do. Never key magenta on warm costumes.
  • Manifest v2 pastes at cell origin; use --fit scale only for legacy sheets that need centring.
  • Pixel art through the aligner: --no-scale, and the scaler uses nearest below 256 px.
  • Turnarounds: cardinals, then diagonals; diagonals come out shorter, so never skip the align step.

Dependencies

Pillow (numpy optional, speeds pixel_scale.py), ffmpeg, ImageMagick 7, genmedia CLI (fal), rembg (stills only), jp2a (optional), GSAP 3 via CDN. Keys: FAL_KEY, RETRODIFFUSION_API_KEY (loaded by scripts/_secrets.py from env or ~/.config/env/secrets.env).

Reference

  • references/provider-matrix.md, references/landscape-2026-09.md — endpoints, prices, what changed
  • references/turnaround.md, references/video-to-spritesheet.md, references/frame-alignment.md, references/chroma-key-transparency.md — pipeline details
  • references/gemini-svg-generation.md, references/pixel-art-svgs.md, references/gsap-timeline-patterns.md, references/ayotomcs-deconstruction.md, references/lil-agents-character-spec.md — SVG and mascot
  • references/ascii-art-techniques.md, references/ascii-animation-components.md, references/animated-ascii-sprites.md — ASCII
  • references/pipelines/open-rebellion.md — project profile pointer
  • tests/test_deterministic.py — run python3 tests/test_deterministic.py

리뷰

아직 리뷰가 없습니다. 사용해 본 소감을 남겨 주세요.

댓글

로그인 없이 쓸 수 있습니다. 사용 후기, 질문, 작동하지 않은 환경 등을 편하게 남겨 주세요.

아직 댓글이 없습니다. 첫 댓글을 남겨 보세요.

보내면 바로 공개됩니다. 이메일 주소나 전화번호 같은 개인 정보는 쓰지 마세요. URL은 2개까지입니다.

0 / 1,000자

이 양식은 reCAPTCHA로 보호되며 Google 개인정보처리방침 및 서비스 약관이 적용됩니다.

같은 저장소의 스킬

Search academic papers, build literature reviews, and synthesize research findings — combines Exa MCP (research_paper category, arxiv filtering) with arxiv-mcp-server for paper discovery, download, and deep analysis. Triggers on academic paper, literature review, research synthesis, arxiv, find papers, scholarly search.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

tdimino/claude-code-minoan412026년 10월 16일 업데이트

Automates browser interactions for web testing, form filling, screenshots, and data extraction. Use when the user needs to navigate websites, interact with web pages, fill forms, take screenshots, test web applications, or extract information from web pages.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

tdimino/claude-code-minoan412026년 10월 16일 업데이트

This skill should be used when creating, auditing, or maintaining AGENTS.md files and Codex CLI configuration for any project—including initializing AGENTS.md for cross-agent compatibility (Codex, Cursor, Copilot, Devin, Jules, Amp, Gemini CLI), generating config.toml or .rules files, scaffolding .agents/skills/, converting CLAUDE.md to AGENTS.md, or auditing existing agent configs for bloat and staleness. Complementary to codex-orchestrator (which executes subagents; this skill creates the config files they consume).

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

tdimino/claude-code-minoan412026년 10월 16일 업데이트

Academic research skill for Biblical Hebrew, Semitic linguistics, cuneiform studies, and comparative Ancient Near Eastern research. Provides Sefaria API for Hebrew Bible, CDLI/ORACC for cuneiform databases, and web discovery via Omnisearch, Exa, Firecrawl, and Obscura for finding scholarly sources across JSTOR, Perseus, Persée, Google Scholar, and academia.edu. Triggers on Hebrew quotes, cuneiform, Sefaria, ANE research, Minoan, search for scholarship, find papers, literature review, scholarly search, academic search, Genesis/Tehom, Ugaritic, Talmudic sources, extract from PDF, OCR academic.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

tdimino/claude-code-minoan412026년 10월 16일 업데이트

Build comprehensive ARCHITECTURE.md files following matklad's canonical guidelines — bird's-eye views, ASCII/Mermaid diagrams, codemaps, invariants, and layer boundaries. Triggers on document the architecture, create ARCHITECTURE.md, map this codebase, architectural overview.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

tdimino/claude-code-minoan412026년 10월 16일 업데이트

Generate physically-based atmospheric scattering shaders — sky domes, planetary atmospheres, LUT-optimized pipelines, depth-aware post-processing. Four modes — sky-dome, atmosphere-post, planet, lut. Triggers on atmospheric scattering, sky shader, sunset rendering, planet atmosphere, Rayleigh scattering, Mie scattering, volumetric sky, sky dome, atmosphere post-processing, aerial perspective, transmittance LUT, sky rendering, realistic sky, planetary rendering.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

tdimino/claude-code-minoan412026년 10월 16일 업데이트

tdimino의 스킬 모두 보기

이 스킬의 문제 신고