> ## Documentation Index
> Fetch the complete documentation index at: https://familyco.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Skills standard

> How skills are stored, discovered, enabled, and used in FamilyCo.

A **skill** is a packaged capability an agent can use — a reusable prompt, a structured output, or a procedure. FamilyCo discovers skills from disk and makes them switchable per company.

## Canonical path

```txt theme={null}
{CODEBASE}/skills/{SKILL_NAME}/SKILL.md
```

Each skill lives in its own directory. The directory name must match the skill slug (lowercase letters, numbers, hyphens only).

## Required rules

* Every skill lives in its own directory.
* `SKILL.md` is mandatory.
* Supporting files may sit beside `SKILL.md`.
* Skills are scanned on app start and on manual rescan.

## Frontmatter

### Required

* `name` — must equal the directory slug.
* `description` — explain what the skill does and when to use it.

### Optional

* `license`
* `compatibility` — note FamilyCo runtime constraints if relevant.
* `metadata` — may include `owner`, `version`, `tags`, `allowedModules`.

## Recommended folder structure

```txt theme={null}
skills/
└─ finance-analyst/
   ├─ SKILL.md
   ├─ examples.md
   ├─ references.md
   └─ templates/
```

Keep `SKILL.md` concise so discovery stays cheap. Put long examples and references in side files; the runtime loads them only when a task truly needs them.

## Example `SKILL.md`

```md theme={null}
---
name: finance-analyst
description: Analyze budget usage, identify spend anomalies, and produce concise reports for FamilyCo budget reviews.
compatibility: FamilyCo server runtime, markdown output, no browser needed.
metadata:
  owner: familyco
  version: 1.0.0
  tags: [budget, finance, reporting]
---

# When to use
Use when an agent needs to analyze token or cost data.

# Inputs
- budget usage rows
- time window
- grouping level

# Outputs
- summary
- anomalies
- recommendations
```

## Enable / disable behavior

* Disabled skills remain discoverable in the UI but cannot be selected for new runs.
* Existing completed runs still reference disabled skills historically.
* If a run requests a disabled skill, the runtime creates an Inbox warning instead of silently substituting.

See the [Skills module](/modules/skills) for the UI flows.