MENSARA.AI · A GOVERNED MIND, IN YOUR ACCOUNT
Sign in
Mensara Create account

MENSARA.AI · SKILLS, APPS AND AUTOMATIONS · MACHINE SURFACE · TEXT/MARKDOWN

---
title: Skills, apps and automations
canonical_url: https://app.mensara.ai/how-it-works/skills-apps-automations
feature: skills-apps-automations
audience: llm-agent
last_verified: 2026-08-11
availability: partly-available
signup_supported: false
interfaces:
  - portal
  - mcp
  - cli
---

# Skills, apps and automations

Distinguishes skills, apps, automations, and connections; defines manifests, principals, triggers, budgets, install review, sandboxing, and disable behavior.

## Concept model

- Skill: procedural instructions interpreted by an agent.
- App: installed software with its own principal and manifest.
- Automation: schedule or event trigger.
- Connection: controlled authorization to an external system.

## Capabilities

- Review declared zones, verbs, connections, triggers, and budgets before enablement.
- Run components under separate principals.
- Pause, revoke, or remove a component and inspect its receipts.

## Prerequisites

- A supported component package or owner-authored skill.
- Human review of the manifest.
- All requested zones, connections, and budgets explicitly granted.

## Authorization

- Effective authority cannot exceed the component principal's grants.
- Scheduled execution does not bypass approval requirements.
- External spend or write actions require declared capability and budget.

## Connection or use flow

1. Inspect manifest and requested authority.
2. Owner approves, narrows, or rejects it.
3. Install with a distinct principal and optional trigger.
4. Run through the access gate and emit a receipt.
5. Pause or revoke when behavior is no longer wanted.

## Refusals and recovery

- Undeclared zone, verb, connection, or spend: hold the operation.
- Disabled component: do not run even if a trigger fires.
- Unsupported package or unsafe sandbox: refuse installation and preserve the review result.

## Ownership and data boundaries

- Removing a skill or app does not delete owner knowledge.
- Component credentials and secrets remain separate from memory files.
- Third-party code does not receive implicit trust from being installed.

## Availability and limits

**Partly available** Baseline skills, apps, and automations exist. Broad third-party app distribution is not open.
- Untrusted third-party apps require stronger sandbox proof before general installation.
- An automation never gains more authority than its configured principal.

## Signup and provisioning

`signup_supported: false`

A human Owner first provisions the Mind. Components are then reviewed and enabled individually; agents cannot self-install or approve their own manifests.

No autonomous signup or provisioning endpoint is documented because none is offered today. An agent must stop and involve the human owner.

## System diagram

```mermaid
flowchart LR
  T[Automation trigger] --> S[Skill instructions]
  T --> A[App principal]
  C[External connection] --> A
  S --> G{Access gate}
  A --> G
  G --> M[Mind operation]
  M --> R[Run receipt]
```

## Runnable example

```text
Automation: weekday morning brief
Trigger: 07:00 local time
Reads: calendar, priorities, waiting-for
Writes: briefs/working only
External actions: none
On failure: record receipt and notify Owner
```

## Related

- [Connections and capture](https://app.mensara.ai/how-it-works/connections-and-capture)
- [Zones and permissions](https://app.mensara.ai/how-it-works/zones-and-permissions)
- [Changes, approvals and receipts](https://app.mensara.ai/how-it-works/changes-approvals-receipts)