Former GitHub CEO Dohmke's New Company Entire Open Sources Skills Cross-Agent Skill Plugin
Entire, a company founded by former GitHub CEO Thomas Dohmke, has open-sourced the Skills plugin, allowing five major programming agents—Codex, Claude Code, Gemini, OpenCode, and Cursor—to read the same development context and achieve seamless task handovers.
Entire had previously open-sourced a CLI tool that automatically records the agent's prompts, reasoning processes, and modified files at each git commit, generating "checkpoint" snapshots stored in a separate git branch. Skills is built on these checkpoints.
Skills currently offers three core functionalities: automatically generating structured summaries during agent handovers, tracing single-line code back to the agent dialogue that generated it, and supporting searches of historical checkpoints by topic/branch/author/time.
Source: Public Information
ABAB AI Insight
Thomas Dohmke led Copilot from an experimental feature to mainstream during his tenure as GitHub CEO from 2021 to 2025. The founding of Entire and the rapid open-sourcing of Skills continue his deep engagement with the developer tools ecosystem. Previously, Entire completed a $60 million seed round in February (led by Felicis, with a valuation of $300 million), making it one of the largest publicly disclosed seed rounds in the developer tools space.
In terms of capital strategy, Entire will focus on cross-agent context management and historical traceability infrastructure, using open-sourced Skills to reduce friction for developers switching between multiple agents while accumulating real development trajectory data to pave the way for future commercialization of an agent orchestration platform. The team of 15 primarily comes from GitHub and Atlassian, possessing strong product and engineering backgrounds.
Similar to emerging AI coding tools like Cursor and Replit exploring context management, as well as the evolution of GitHub's own Copilot Workspace, Entire is currently in the early infrastructure stage of transforming programming agents from single tools to multi-agent collaboration and traceable workflows.
Essentially, this represents a technological substitution: Skills reconstructs dispersed agent dialogues and code changes into a unified traceable knowledge layer through the checkpoint mechanism, shifting capital from repetitive manual context transfer to automated summarization and search. Mechanically, it lowers the adoption threshold through native git integration, accelerating the structural shift of AI programming from "single-agent assistance" to a "multi-agent collaborative operating system."
ABAB News · Cognitive Laws
The best developer tools do not make a single agent smarter, but enable multiple agents to truly pass the baton. Each code change should record its "birth certificate," which is the true foundation of trust in the AI programming era. When open source becomes the standard for the largest seed rounds, the competition in developer tools has entered an infrastructure battle.