datajack 0.1.0

The command-line surface of the skillwire package: a `skill` module of five routes that any modular_cli_sdk CLI mounts to deploy agent skills into AI coding hosts. The port a person plugs into; skillw [...]

datajack

The command-line surface of the skillwire package: a skill module of five routes that any modular_cli_sdk CLI mounts to deploy agent skills into AI coding hosts.

In Shadowrun a datajack is the port through which a person connects to a device. The wire moves the skills; this is where a human plugs in to govern it.

Mount it

cli.module('skill', (m) => buildSkillModule(
  m,
  consumer: 'macss',      // written into every ledger row this CLI creates
  workspace: workspace,
  catalogue: catalogue,
));

That is the whole integration. consumer is what makes a shared machine work: the ledger records who deployed what, so a second CLI meeting your artifact reports it as owned rather than overwriting it.

The five routes

Route	Kind
skill list	Query	Catalogue and status in one table, with what else can see it
skill deploy	Command	Reconcile a host toward the skills this release ships
skill remove	Command	Reconcile away from them, touching only what this consumer deployed
skill doctor	Query	What is deployed, who owns it, what has drifted
skill validate	Query	Conformance to the Agent Skills specification

Every Command requires --plan or --apply; neither is a default. --host and --scope are required and have no defaults either.

Why a separate package

Because it is irreducibly a command-line surface, and the domain is not. It declares --host and --scope, maps failures onto exit codes, and renders an aligned table for a terminal — none of which survives a move to another transport, and none of which belongs in a library about reconciling directories.

Keeping them apart also keeps skillwire off the SDK's release cadence: were the domain to depend on modular_cli_sdk, it would become the gate through which every consumer reaches a new SDK version.

See ADR 0005.

---

Built with MACSS.