Architecture

This repo's architecture follows one defining constraint: shipped skills run with no install step. The runtime files that provider manifests, docs, and users execute are committed, dependency-free Node ESM that use only the standard library, with the owned provider CLI as the single external boundary.

To keep those committed runtime files maintainable, the source of truth for the non-trivial ones lives in canonical TypeScript under src/. That TypeScript is developer tooling, not a runtime dependency: a build step compiles it into the committed .mjs outputs that ship. TypeScript, Vitest, and bundling never become a runtime dependency of a shipped skill.

Three pieces of this architecture have their own pages:

• Shared transcript-core — per-provider transcript knowledge has one source of truth, and each consuming skill ships a committed generated copy rather than importing across skills.
• Cursor collaboration reliability — exact identity, physical-frame observation, isolated continuity state, and lease-scoped completion keep availability separate from automatic continuation.
• Generated runtime outputs — the build contract that maps canonical TypeScript sources to their committed .mjs outputs, and the rule that those outputs are never hand-edited.

Contents

• Shared transcript-core — One source of truth for per-provider transcript knowledge, with a committed generated copy per consuming skill.
• Cursor collaboration reliability — Exact identity, content-first observation, state continuity, and lease-scoped completion for Cursor collaboration.
• Generated runtime outputs — The canonical-TypeScript → committed-.mjs build contract and the never-hand-edit rule.