What is omnistack-agent?

omnistack-agent is a collection of ready-to-use instruction files that tell an AI assistant to act as a full-stack software engineering specialist. Rather than writing your own system prompt from scratch, you copy one of the pre-built adapter files and paste it wherever your chosen AI platform expects its instructions. The idea is that one set of source instructions is compiled into separate files for each supported platform: ChatGPT custom GPTs, Claude skills and agents, GitHub Copilot, Google Gemini, Cursor and Windsurf editors, and any generic AI via a plain system prompt. You pick the file that matches your tool and paste it in. No installation or dependencies required. Once active, the agent can take on ten different engineering roles depending on what a task needs: software architect, full-stack developer, mobile developer, backend engineer, frontend engineer, database administrator, DevOps engineer, QA engineer, technical writer, or software mentor. It shifts between those roles based on what you ask, rather than being locked into one specialty. The source lives in two folders. The core folder holds the agent's identity, principles, and guardrails. The knowledge folder holds modular topic files, one file per subject. Adapter files are generated from those sources using a zero-dependency Node.js script. You should never edit the adapter files directly, because the build command regenerates them and would overwrite any hand edits. If you want to add a new topic or change the agent's behavior, you edit the source files and run the build step to produce updated adapters. Contributing follows a clear workflow: edit the source, run the build, run the tests, and open a pull request. The CI pipeline checks that committed adapter files match the source, so a pull request fails if they drift out of sync. The project requires Node 18 or newer and carries no npm dependencies. It is released under the MIT license.