Skip to main content

Workflows

Short, task-oriented sequences for the things you actually do with STAK. Each one names the real command — for every option and flag, see the CLI Reference.

See what changed

Read-only — compares this machine against main. Nothing moves yet.

stak status

Choose what belongs in main

Stage the Skills and Memory you want to keep in main.

stak add .

Sync it

Synchronize staged main Skills and Memory to STAK Cloud.

stak push

Discover existing Skills

A diagnostic, read-only discovery across agent Skill roots — not where the daily workflow starts (that's stak status).

stak scan

Create main

Capture the Skills currently installed globally as your desired Environment, main. Every later run, status, or diff compares against this.

stak init

Use a STAK

Add a STAK layer — local path or public @owner/name — to main. This only updates desired state; it does not launch an agent.

stak use @owner/name

Try a STAK safely

Run a STAK in an isolated, temporary runtime without touching main, your normal agent bindings, or any permanent agent state. The trial runtime is cleaned up afterward.

stak try @owner/name -- codex

Run an agent

Reconcile main into an agent's Skill directory and launch that agent's CLI with the Skills projected.

stak run claude

Publish a STAK

Publish a public STAK as @owner/name from a local source or an active local layer.

stak publish ./my-stak --slug my-stak

Update a published STAK

There is no automatic update in V0.1 — an already-active public STAK is never silently upgraded. Check each Skill's source for newer versions, then re-run use to pick them up.

stak skill check
stak use @owner/name

Back up Memory

Copy allowlisted Memory into STAK's private store. This is a read-only capture — native agent memory files are never modified. This is the explicit, manual backup command — the normal stak push flow already syncs staged Memory; use this when you want an out-of-band backup.

stak memory backup

Recover Memory

stak memory recover downloads your private memory store from STAK Cloud; stak memory restore then writes it into native agent Memory on this machine. Existing targets are never overwritten — a differing existing target is reported as a conflict and left untouched. This is explicit, history-based recovery — not part of the normal new-machine flow, which uses stak pull instead (see below).

stak memory recover
stak memory restore

Map recovered Projects

Recovered Memory surfaces as logical Projects. Review them, then authorize a local directory as the restore destination for one — this is a local binding only, never shared as part of the Project's identity.

stak project
stak project map my-app ~/code/my-app

Move to another machine

On a new machine, sign in and pull — this restores main, Skills, and your latest staged private Memory generation:

stak login
stak pull
stak status

Pull does not touch your agents, so stak status lists the pulled Skills as not installed here yet. Set them up in an agent:

stak run codex

If you use Claude Code, materialize recovered Memory into it:

stak project
stak memory restore --dry-run
stak memory restore

stak memory recover remains available for explicit or history-based recovery (see CLI Reference) but is not part of this normal flow.