Mental Models
Syllago borrows ideas from package managers, virtual environments, build tools, and version control. If a word feels unfamiliar, the chances are good you already have a working mental model from somewhere else.
Vocabulary cheat sheet
Section titled “Vocabulary cheat sheet”| Syllago term | Familiar analog | Why it’s similar |
|---|---|---|
| Registry | npm registry, Homebrew tap, apt source | A networked source of installable content that someone else maintains. |
| Library | ~/.npm/, ~/.cargo/, your local dotfiles | The on-machine store of content you own and edit. |
| Loadout | Python venv, docker-compose.yml, a Terraform workspace | A named bundle you apply as a unit and revert cleanly. |
| Provider | A deploy target (Vercel, Netlify, Heroku) | The thing your content ships to — --to <provider> is the destination. |
| Content type | A package or artifact format | The kind of thing (rule, skill, agent, hook…). Each type has its own conventions. |
| Canonical key | An interface, or a portable IR | The tool-neutral name for a capability (e.g. hierarchical_loading) that providers implement differently. |
| Format conversion | A transpiler (Babel, esbuild target) | What syllago does between its internal representation and each provider’s native format. |
| Hub-and-spoke model | A compiler’s intermediate representation, GraphQL between front and back ends | One neutral middle format; providers connect through it instead of converting pairwise. |
| Privacy gate | A .gitignore for AI content | A guard that stops marked-private content from leaking to public registries. |
| MOAT trust tier | Sigstore, npm provenance attestation | The signed-and-verified provenance level of a registry. |
| Sandbox | Docker, bubblewrap, an unprivileged user | A confined environment for running an AI provider with restricted FS, network, and env access. |
When you’d reach for each one
Section titled “When you’d reach for each one”- Trying things out on your own machine? Library +
install. - Need a coworker to use the same content? Registry (or commit
.syllago/to the repo — see Local vs. Global). - Want to switch between a “Python backend” and a “frontend review” context? Loadout.
- Targeting a tool other than the one you authored in? Format conversion runs automatically on install.
- Asking “will this work in Kiro?” Run
syllago compat.
Related
Section titled “Related”- Why syllago — the problem these models exist to solve.
- Core Concepts — the formal definitions.
- Format Conversion — how the hub-and-spoke model works in practice.