Recommended Skills Library

Skills are reusable playbooks loaded into an agent's working context when a specific task requires domain expertise. Think of them as expert checklists that encode institutional knowledge: the patterns a senior engineer would follow, the edge cases a QA lead would catch, the style conventions a tech writer would enforce. They keep the main context window lightweight by loading on-demand, and they ensure repeatable quality regardless of which team member (or agent) executes the task.

The Skills Ecosystem in 2026 #

The primary registry for discovering and installing skills is skills.sh, which hosts thousands of community and vendor-published skills with security assessments. Installation is a single command:

# Install a skill globally (available across all projects)
npx skills add <author>/<repo>@<skill-name> -g

# Install a skill locally (scoped to current project)
npx skills add <author>/<repo>@<skill-name>

Skills are stored as directories containing a SKILL.md file (Markdown with YAML frontmatter) and optional supporting resources (scripts, examples, references). Most modern harnesses—Antigravity, Claude Code, Cursor, Cline—support loading skills from .agents/skills/ (project-scoped) or ~/.agents/skills/ (global).

Recommended Skills by Category #

Planning & Alignment #

Code Quality & Simplicity #

Writing & Documentation #

UI Design & Craft #

Research & Investigation #

Automated Review Agents #

Writing Custom Skills #

When your team develops domain-specific expertise that no existing skill covers—your proprietary deployment pipeline, your industry-specific compliance checks, your internal API conventions—encode it as a custom skill rather than relying on individual engineers to remember and enforce the pattern manually.

A minimal custom skill requires:

.agents/skills/your-skill-name/
└── SKILL.md

The SKILL.md file uses YAML frontmatter for metadata and Markdown body for instructions:

---
name: your-skill-name
description: One-line description of what this skill does and when to trigger it.
---

# Your Skill Name

## When to Use
Describe the trigger conditions.

## Instructions
Step-by-step process the agent should follow.

## Constraints
Hard rules the agent must not violate.

## Output Format
What the agent should produce when done.

For more complex skills, add supporting directories: scripts/ for helper scripts, examples/ for reference implementations, references/ for additional documentation, and resources/ for templates or assets.

Cross-Harness Compatibility #

The SKILL.md convention originated in Claude Code and has been adopted by most major harnesses, though with varying levels of native support: