Agent Skills

How skills work

AI agents are increasingly capable, but they often lack the specific context needed for real work. Without knowledge of your team's deployment process, your company's code review standards, or your project's data analysis pipeline, agents guess and iterate.

Loading all this context upfront isn't practical either. Too much information overwhelms the agent, slowing responses and reducing quality.

Skills solve this with progressive disclosure:

1. Discovery — At startup, Mind loads only the name and description of each skill
2. Activation — When your request matches a skill's description, Mind loads the full instructions
3. Execution — Mind follows the instructions, loading scripts or reference files only as needed

Using skills

Skills can activate in two ways:

• Automatically — Mind matches your request against skill descriptions and loads the relevant skill
• As slash commands — Type / followed by the skill name to invoke it directly

Passing arguments to a skill

If the skill body contains $ARGUMENTS or $ placeholders, text after the slash command is substituted into them:

Skill scope

When skills share the same name, workspace skills take priority over global skills.

Custom agents and skills

{
  "name": "my-agent",
  "resources": [
    "skill://.mind/skills/*/SKILL.md",
    "skill://~/.mind/skills/*/SKILL.md"
  ]
}

Creating a skill

A skill is a folder containing a SKILL.md file:

my-skill/
├── SKILL.md           # Required
├── scripts/           # Optional executable code
├── references/        # Optional documentation
└── assets/            # Optional templates

SKILL.md format

---
name: pr-review
description: Review pull requests for code quality, security issues, and test coverage. Use when reviewing PRs or preparing code for review.
---

## Review checklist

When reviewing a pull request:

1. Check for vulnerabilities, injection risks, exposed secrets
2. Verify edge cases and failure modes are handled
3. Confirm new code has appropriate tests
4. Ensure variables and functions have clear names

## Common issues to flag

- Hardcoded credentials or API keys
- Missing input validation
- Unhandled promise rejections
- Console.log statements left in production code

Frontmatter fields