Teach Imaginne how your company works.
A skill packages one of your organization's procedures — the process, the instructions, the tools and the credentials it needs — and turns it into something any authorized person triggers by conversation.
What fits in a skill
Process
The step-by-step that lives in someone's head today: how the monthly report is produced, how an environment is provisioned, how an internal system is operated.
Knowledge
The rules, standards and context that make the result come out your company's way — not the generic way.
Tools
Scripts and integrations the skill uses to actually execute: APIs, databases, cloud, spreadsheets.
Credentials
The secrets execution requires, stored in the organization vault — never in the skill's text, never in the conversation.
How it works day to day
For the people using it
You ask for the outcome in natural language — "generate the closing report" — and Imaginne triggers the right skill. You don't need to know the skill exists, or how it works inside.
For the people publishing
The organization publishes skills to its profiles: who can use each one is decided by policy, not by informal arrangement. Publish once, everyone authorized has it.
For IT
Activation is explicit and auditable. The skill declares which secrets it needs; the platform resolves credentials only at execution time, scoped per step, and refuses to publish if a binding is missing.
- Skill "monthly-closing" triggered
- Data collected from connected systems
- Report generated in the company standard
- E-mail sent to the board
Why this matters
Operational knowledge stops depending on each person explaining it again. The procedure becomes an organizational asset: versioned in its bundle, distributed by profile, executed the same way every time — in an isolated environment, with credentials managed by the platform.
Technical details
Skills are bundles published by the organization through an administrative API. Execution happens in Imaginne's isolated environment (kernel-isolated sandbox); on Desktop and Terminal, suitable skills can execute locally. The skill↔secret contract is verified at publish time: declared secrets must be bound, and a referenced secret cannot be deleted. Activation in a conversation is explicit — the model does not load every skill body in the organization, only what the user's policy allows.