upgradeables

Upgradeables is a repository of reusable, model-agnostic building blocks for AI work. These building blocks, called Upgradeables, are optional capabilities with defined triggers, inputs, outputs, procedure, compatibility rules, failure boundary, and tests. They can be used directly in a chat, combined into a reusable Skill, or adapted to an agent system. The repository provides multiple entry points, including a human/model router (START_HERE.md), a tiny machine router (runtime/router.json), and compact runtime cards and recipe packs. Users can start in 30 seconds by pasting a prompt into a model that can open web links, or by cloning the repository for IDE or agent use. The project also includes a local-first harness with no runtime dependencies, installable via pipx, which allows initializing and integrating Upgradeables into a project. The repository emphasizes that host and system policy always win, and Upgradeables cannot provide hidden memory, private chain-of-thought access, unavailable tools, or safety bypasses.