An AgentSkill that turns GitHub's contribution graph into a pixel art canvas.
Tell your AI agent to paint pixel art on your GitHub profile. It handles the design — the skill handles the git plumbing.
Your GitHub contribution graph is a 52×7 grid of green squares. This skill teaches AI agents how to paint on it by creating backdated commits to a private repo. The art shows up on your profile.
Drop this folder into your project or global skills directory:
# Claude Code / Cursor / VS Code
cp -r painting-github-heatmaps .agents/skills/
# Or clone directly
git clone https://github.com/jmzlx/github-heatmap-art .agents/skills/painting-github-heatmapsThe script auto-detects your git config and gh CLI credentials. If you have git configured and gh auth login done, zero setup needed.
Optional overrides:
export GITHUB_TOKEN="ghp_..." # if no gh CLI
export GIT_AUTHOR_NAME="Your Name" # if no git config
export GIT_AUTHOR_EMAIL="12345+you@users.noreply.github.com" # noreply recommended
export HEATMAP_REPO="heatmap-art" # default worksThen ask your agent:
"Paint a heart on my 2020 GitHub heatmap"
- Agent composes a 7×52 grid of intensities (0–4) — it writes code to build the grid
- Agent previews as ASCII art — verifies the design looks right
- Script renders —
scripts/paint.shcreates backdated commits and pushes in batches
The agent handles all creative work (design, sprites, text, patterns). The script handles all git plumbing (backdating, batching, year boundaries).
Works with any AgentSkills-compatible tool:
Claude Code · Cursor · Gemini CLI · VS Code Copilot · GitHub Copilot · OpenClaw · Goose · OpenHands · Amp · Junie · OpenCode · Mux · Firebender · Letta · Autohand
painting-github-heatmaps/
├── SKILL.md # Agent instructions (<5000 tokens)
├── scripts/paint.sh # Git plumbing (batch commits + push)
├── references/design-guide.md # What works on real heatmaps
└── README.md # This file
This was originally an MCP server with 12 typed tools. We killed it. A coding agent doesn't need structured tools to create a 7×52 array — it just needs to know the domain and have a script for the git plumbing. The skill approach is simpler, works with more agents, and costs zero context tokens until you actually need it.
MIT