Familiar logo
Familiar usefamiliar.com
Documentation / Fleet

Skills

Build reusable skills that give agents specialized capabilities.

What Are Skills?

Fleet skills management

Skills are portable packages of instructions, scripts, and resources that agents can use to accomplish tasks. They follow the open Agent Skills specification.

A skill can contain:

  • Instructions — Step-by-step procedures in markdown
  • Scripts — Executable code the agent can run
  • Assets — Templates, reference files, and resources

Skill Structure

Each skill is a folder containing a skill.md file and supporting assets:

my-skill/
├── skill.md          # Instructions and metadata (required)
├── template.html     # Template file (optional)
├── process.sh        # Script (optional)
└── examples/         # Reference materials (optional)

skill.md Format

Skills use YAML frontmatter for metadata, followed by markdown instructions:

---
name: code-review
description: Reviews code for quality, security, and maintainability
version: 1
compatibility: all
is_command: true
provider: anthropic
model: claude-sonnet-4-20250514
tier: read_only
allowed_tools:
  - Read
  - Glob
  - Grep
---

# Code Review

Review code for quality, security, and maintainability.

## Steps

1. Read the code completely before commenting
2. Check for logical errors and bugs
3. Look for security vulnerabilities
...

Frontmatter Fields

Field Type Required Description
name string Yes Unique skill identifier
description string Yes Brief description for the skills browser
version integer No Skill version number
compatibility string No Agent compatibility: all, cli, standard
is_command boolean No If true, invokable with /skill-name
provider string No Preferred AI provider
model string No Preferred model
tier string No Suggested execution tier
allowed_tools string[] No Tools the skill expects to use (guidance only)
is_core boolean No True for built-in core skills

Tool Guidance

The allowed_tools field is guidance for the model—it's not enforced. Actual tool access is controlled by the agent's execution tier and permission rules.

Skill Discovery

Fleet automatically discovers skills from:

Default location:

  • macOS: ~/Library/Application Support/familiar/skills/
  • Linux/Windows: ~/.familiar/skills/

Fleet scans recursively for skill.md files:

skills/
├── development/
│   ├── code-review/
│   │   └── skill.md
│   └── pr-summary/
│       └── skill.md
├── writing/
│   ├── blog-post/
│   │   └── skill.md
│   └── email-draft/
│       └── skill.md
└── data/
    └── csv-analysis/
        └── skill.md

Skills are deduplicated by name—if two skills have the same name, only one loads.

Creating Skills

Option 1: Ask an Agent (Recommended)

Ask any agent to create a skill:

Create a skill called "Code Review" that reviews code for bugs,
security issues, and performance problems.

The agent creates the skill folder and writes the skill.md file.

Option 2: Create Manually

  1. Open the skills folder: Settings → Skills → Open Folder
  2. Create a new folder (e.g., my-skill/)
  3. Create skill.md with frontmatter and instructions
  4. Add supporting scripts or templates

Option 3: Skills Browser

  1. Open Settings → Skills or the Skills panel
  2. Click New Skill
  3. Fill in name and description
  4. Edit content in the editor
  5. Save

Using Skills

Agents automatically discover and use relevant skills. To explicitly request a skill:

Use the Code Review skill on src/api/auth.ts

Enable/Disable Skills Per Agent

Control which skills an agent can access in AGENT.md:

enabled_skills:
  - code-review
  - documentation
  # Empty array means all skills enabled

Command Skills

Skills with is_command: true can be invoked directly:

/code-review src/api/

Fleet translates this into a skill invocation with the argument.

Importing Skills

From GitHub

Ask an agent to import from a skill repository:

Import the code-review skill from github.com/example/skills

Manual Import

  1. Clone or download the skill repository
  2. Copy the skill folder to your skills directory
  3. Fleet discovers it automatically

Example: Code Review Skill

---
name: code-review
description: Reviews code for quality, security, and maintainability
version: 1
compatibility: all
is_command: true
tier: read_only
allowed_tools:
  - Read
  - Glob
  - Grep
---

# Code Review

Review code systematically for issues.

## Review Checklist

### 1. Security (Critical)
- [ ] No hardcoded secrets or credentials
- [ ] Input validation on external data
- [ ] No injection vulnerabilities
- [ ] Proper authentication/authorization

### 2. Bugs
- [ ] Null/undefined properly handled
- [ ] Edge cases covered
- [ ] Error handling complete

### 3. Performance
- [ ] No unnecessary loops
- [ ] Database queries optimized
- [ ] Appropriate caching

### 4. Maintainability
- [ ] Clear naming
- [ ] Reasonable function sizes
- [ ] Minimal duplication

## Output Format

For each issue:
1. **Location:** File path and line number
2. **Severity:** Critical / High / Medium / Low
3. **Issue:** What's wrong
4. **Fix:** How to correct it

Managing Skills

Skills Browser

Access via Settings → Skills or the sidebar.

Features:

  • View all installed skills
  • Search by name or description
  • Edit skill content
  • Delete skills
  • Open skills directory

Columns

Column Description
Name Skill identifier
Description What the skill does
Version Skill version number
Path Location on disk

Click headers to sort. Resize by dragging.

Skill Fields Reference

Complete list of skill.md frontmatter fields:

Field Type Description
name string Unique identifier
description string Brief description
version integer Version number
is_command boolean Invokable with /name
provider string Preferred provider
model string Preferred model
tier string Suggested execution tier
compatibility string Agent type compatibility
allowed_tools string[] Expected tools (guidance)
is_core boolean Built-in core skill

Next Steps