GitHub Action for Measuring Codebase Size in Tokens with Context Window Badge
By
jimminyx
Crisp on the outside, thoughtful on the inside. A keeper.
Summary
This article describes nanoclaw/repo-tokens, a GitHub Action that calculates codebase size in tokens and updates a README badge with color coding based on what percentage of an LLM's context window the codebase fills. The tool uses tiktoken for token counting and provides visual feedback (green, yellow-green, yellow, red) for different percentage ranges. The article explains that small codebases are particularly advantageous with coding agents, and mentions that nanoclaw is a lightweight alternative to OpenClaw that runs in containers for security and connects to various messaging platforms.
Key quotes
· 4 pulledThis counts tokens using tiktoken and writes the result between HTML comment markers in your README
The badge color reflects what percentage of an LLMs context window the codebase fills (context window size is configurable, defaults to 200k which is the size of Claude Opus)
Small codebases were always a good thing. With coding agents, there's now a huge advantage to having
A lightweight alternative to OpenClaw that runs in containers for security. Connects to WhatsApp, Telegram, Slack, Discord, Gmail and other messaging apps, has memory, scheduled jobs
You might also wanna read
Researcher's "ADHD" tool for Claude Code claims 2x improvement; experts call for more evidence
Solo researcher Udit Akhouri released a third-party Agent SDK tool called "ADHD" for Claude Code on Reddit, claiming it helps coding agents
bit.ly·9h ago
Microsoft Foundry Skill: Reusable Guidance for Coding Agent Workflows
The Microsoft Foundry Skill is a meta-skill for coding agents that provides reusable guidance for Foundry agent workflows. It helps standard
Erlang Ecosystem Foundation Launches Central Company Registry
The Erlang Ecosystem Foundation has launched a new canonical company registry at registry.erlef.org. The registry aims to serve members, spo
aislop: Open-source CLI tool detects low-quality code patterns left by AI coding agents
aislop is an open-source CLI tool (MIT-licensed) that detects low-quality patterns left behind by AI coding agents like Claude Code, Cursor,
github.com·2d ago
wterm: A DOM-based Web Terminal Emulator Powered by Zig and WebAssembly
wterm is a web-based terminal emulator that renders directly to the DOM, providing native text selection, copy/paste, find functionality, an
wterm.dev·2d ago
PromptLayer: AI observability tool for tracing multi-step workflows, costs, and failures
PromptLayer is an AI observability tool for developers that provides a unified timeline and waterfall view to trace AI requests, workflows,
Product Hunt·3d ago