agentscan

AgentScan is an open-source tool designed for open source maintainers to detect automation patterns in GitHub accounts. It analyzes a GitHub account's public activity and scores it based on how much it appears to rely on automation, without using AI—just event analysis powered by the identity library. The results are presented as indicators, not verdicts, and are meant as a starting point for maintainers' own judgment. The project offers two integration methods for using the same analysis directly in your own repository: - **GitHub App** (zero-config): Installs and works immediately, updates centrally. - **GitHub Action**: Requires adding to a workflow and manually pinning versions. It also maintains a curated list of manually verified automated accounts, submitted by the community and reviewed by maintainers before being merged. Users can report suspected automated accounts via a template issue, and flagged accounts can dispute classifications through linked issues. The hourly Ecosystem Activity scan runs in a separate repository (agentscan-logs), which scores pull requests across tracked repos and serves results over HTTP. This repository hosts the website, GitHub App, webhook, and verified automations list. Stack: Nuxt, UnoCSS, Netlify, GitHub API. Built in response to maintainers facing AI-agent-driven pull requests targeting open source projects. The project is an ongoing experiment, with scores acknowledged as potentially inaccurate and intended as a starting point, not a conclusion.