An Electron-based prompt launcher and knowledge-base host that bootstraps stateless AI browser windows with the exact context and capabilities they need — no more, no less.
Every time I open a fresh AI browser tab — ChatGPT, Claude.ai, whatever — it starts stateless. It doesn't know my projects, my conventions, the shape of the code I'm working in, or which capability primer applies to the task at hand. Every session pays that cost from scratch.
PromptSM is my daily-driver app for making the stateless case pleasant. A Markdown-based library of prompts organized by project and topic; a keystroke that renders the right one to the clipboard; a chord system for dispatching asks to a corporate chat API; a bundled knowledge base of capability primers so an AI window can pull just the context it needs for the task in front of it.
%APPDATA%. Edits propagate to the overlay so a running session can pick them up without a rebuild — and a corrupted overlay can be nuked back to the repo baseline in one command.kb list, kb search, kb show verbs let an AI window discover and pull primers on demand. The primers themselves are Markdown, versioned in the repo.ask <prompt> — renders a prompt from the library and dispatches it to a chat API.apply — syncs repo changes to the %APPDATA% overlay and restarts the chord server without losing state.kb {list,search,show} — discovery + retrieval on the primer set.