Skillbase Sync

What is Sync?

Skillbase Sync keeps your team's AI agents on the same page. It solves two everyday problems:

Everyone configures agents differently. One developer uses five skills, another uses two different ones, a third forgot to install anything. Sync lets the team lead define the project's skill set once — everyone else installs with one click.

Context dies between sessions. You spend an hour working with an AI agent, make important decisions, discover constraints — and all that context disappears. When a teammate picks up the same task, their agent starts from scratch. Sync captures structured knowledge as you work and makes it available to the next person automatically.

Quick start

1. Sign in and create a company

Go to sync.skillbase.space, sign in with GitHub, and create a company. Invite your teammates.

2. Connect your local SPM

On the company page, click Connect. This links your local environment to the company — a one-time setup that takes a few seconds.

3. Bind a project to your repo

spm sync init

Pick a project (or create one), and spm writes a .skillbase/sync.json into your repo. Commit it — now every teammate who clones the repo is bound to the same Sync project.

4. Sync skills and personas

Add skills and personas to the project in the web UI. Then:

spm sync

Missing packages are installed, everything already up to date is skipped. You can also click Sync in the web UI — the SPM daemon handles installation in the background.

Projects

A project in Sync maps to a codebase. It holds:

• Skills and personas — the team's standard toolset, referenced from the registry. Everyone syncs the same set.
• Project prompt — shared context for all agents working on the project. Think of it as a team-wide CLAUDE.md: project overview, tech stack, conventions. Editable in the web UI.
• Features — units of work with structured knowledge (see below).

Features and knowledge

A feature represents a task, ticket, or initiative. As your AI agent works on a feature, it captures knowledge — not as free text, but as typed items:

• Facts — "gray-matter is CJS-only and incompatible with Turbopack"
• Decisions — "Use js-yaml instead" + why: "Need pure ESM for Turbopack pipeline"
• Constraints — "API response must be under 100KB — mobile data budget"
• Artifacts — "Created proto/service.proto — gRPC schema for internal services"
• Open questions — "Backward compat for mobile clients — need PM input"

Why types matter: an agent can ask "what constraints exist?" and get a precise answer. With a free-text wiki, that question turns into guesswork.

When a teammate picks up the same feature, their agent loads all of this automatically — decisions with reasons, unresolved questions, constraints that took you an hour to discover.

Knowledge update modes

Each project has a mode that controls how agents save knowledge:

• Auto (default) — the agent saves knowledge immediately as it discovers things. Fast, zero friction.
• Confirm — the agent proposes what to save and waits for your approval. Good for teams that want human review.

Feature lifecycle

Features move through statuses: draft → active → review → done → archived. Every knowledge update bumps the version, and the full history shows who changed what, when, and from which client.

How agents use Sync

Once connected, everything is automatic — no manual steps during work:

1. You mention a feature ("working on auth-refactor") — the agent loads its context
2. The agent sees existing decisions, constraints, open questions — starts with full picture
3. As it works, it saves new discoveries (in auto mode) or proposes them (in confirm mode)
4. During long sessions, the agent checks for updates from teammates — if someone added a constraint, it adjusts

The agent gets 8 MCP tools when Sync is connected: loading features, pushing knowledge, checking for teammate updates, searching across the project, and more. You don't need to invoke them — the agent uses them based on context.

CLI commands

spm sync

Install missing skills and personas from the project manifest.

spm sync                          # Sync current project (from .skillbase/sync.json)
spm sync backend-api              # Sync a specific project by slug
spm sync --status                 # Dry run — show what's missing without installing
spm sync --company acme           # Sync all projects in the company

spm sync init

Bind the current directory to a Sync project.

spm sync init

Interactive setup: pick a company, pick or create a project, optionally link the git remote. Creates .skillbase/sync.json — commit it so the whole team shares the binding.

Team roles

Search

Sync indexes all features and knowledge items with full-text search. Search from the web UI, via the sync_search MCP tool, or through the API — across a single project or the entire company.

Next steps

• CLI Reference — all spm commands including sync
• MCP Integration — how spm connects to AI clients