Agent Skills — AI Agent Skill Collection

Autonomous skill instructions for AI agents: Sisyphus, opencode, and compatible. Each skill is a folder with SKILL.md (instructions) and skill.json (manifest for installation/discovery).

🌐 Versions: English · Русский · Website

• 📦 Skill Catalog — 42 skills · 5 categories
• 🎬 Showcase — real-project examples
• 🚀 Installation
• 📁 Repository Structure
• 🔍 Skill Discovery
• 🛠 Adding a New Skill
• 📄 Project Info — license · contributing · security · code of conduct

📦 Skill Catalog

📚 Ecosystem map: see docs/SKILLS_CATALOG.md — catalog of the AI agent skills ecosystem (official vendors, formats, standards, ~4.85M SKILL.md files tracked, gaps and recommendations).

42 skills · 5 categories. Pick a category below; triggers and full metadata live in index.json.

💻 Development & Code · code · 14

| long-running-agent-workflow | Protocol for AI agents working across many context windows: a .lra/ directory with an atomic feature list (id/priority/acceptance criteria/status) and session progress checkpoints. |

📊 Data & Diagrams · data · 6

🎬 Content & Media · media · 12

🏗️ Repository & Docs · repository · 10

💬 Social · social · 1

🎬 Showcase — real-project examples

Live demonstrations of skills on real projects (not abstract examples). Reference project: lovii.ru (lovii_demo). Every new skill must ship a showcase. Template: docs/showcase-template.md.

🚀 Installation

For opencode

Copy the desired skill folder to ~/.config/opencode/skills/:

# Example: install presentation-maker
cp -r skills/presentation-maker ~/.config/opencode/skills/

Or load directly via skill tool pointing to SKILL.md:

# In opencode session
skill load path/to/skills/presentation-maker/SKILL.md

For Sisyphus / other agents

Each skill contains:

• SKILL.md — full instruction (Markdown)
• skill.json — manifest with metadata (name, version, triggers, requirements, files)

Agent can parse index.json to discover skills by triggers/categories and load the needed one.

📁 Repository Structure

agent-skills/
├── index.json                 # Master catalog: name, version, category, description, triggers
├── README.md · README.ru.md   # This document (EN / RU mirror)
├── CHANGELOG.md · LICENSE · CONTRIBUTING.md · SECURITY.md · SUPPORT.md · CODE_OF_CONDUCT.md · FUNDING.yml
├── og-image.png               # Social preview 1280×640
├── assets/                    # README header/footer SVG
├── docs/
│   ├── SKILLS_CATALOG.md      # AI agent skills ecosystem catalog
│   ├── showcase/              # Real-project demonstrations (+ template)
│   └── plans/                 # Planning documents
├── .github/
│   ├── ISSUE_TEMPLATE/ · pull_request_template.md · release.yml
│   └── workflows/validate-skills.yml   # CI: manifest validation
└── skills/                    # 42 skills, one folder each
    └── <skill-name>/
        ├── SKILL.md           # Agent instruction (required)
        ├── skill.json         # Manifest: name, version, triggers, files (required)
        ├── scripts/           # Executable helpers (Python/JS)
        ├── templates/         # Reusable templates
        └── references/        # Deep-dive materials

Skill folder anatomy (example — code-review):

skills/code-review/
├── SKILL.md              # Agent instruction: intro → steps → examples → constraints
├── skill.json            # Manifest: name, version, category, triggers, files
├── scripts/              # review.py, checklists.py
├── templates/            # review-template.md
└── examples/             # example-pr.md

Only SKILL.md and skill.json are required — everything else is optional and depends on the skill.

🔍 Skill Discovery

Use index.json — it contains a skills array with fields:

• name, version, category, description, path, triggers, updated

Example filtering by trigger (Python):

import json
with open('index.json') as f:
    data = json.load(f)
# Find skills matching trigger "presentation"
matches = [s for s in data['skills'] if 'presentation' in ' '.join(s['triggers'])]

🛠 Adding a New Skill

1. Create folder in skills/<name>/
2. Add two required files:
   • SKILL.md — full agent instruction (English primary; Russian optional, with YAML frontmatter name, description)
   • skill.json — manifest (see schema below)
3. Optionally add scripts/templates in subfolders (scripts/, templates/, icons/)
4. Update index.json (add entry to skills[])
5. Open PR

skill.json Schema (required fields)

{
  "name": "kebab-case-name",
  "version": "1.0.0",
  "description": "Brief description (1-2 sentences)",
  "author": "github-username",
  "license": "MIT",
  "keywords": ["tag1", "tag2"],
  "triggers": ["trigger phrase 1", "trigger на русском"],
  "category": "repository|media|social|code|data",
  "entrypoint": "SKILL.md",
  "files": ["SKILL.md", "scripts/*.py"],
  "requirements": {
    "tools": ["python3", "gh"],
    "permissions": ["repo:read"]
  },
  "updated": "YYYY-MM-DD"
}

Categories: repository, media, social, code, data

Triggers: phrases that cause the agent to load the skill. Provide English; Russian optional.

SKILL.md Requirements

• Language: English (primary, instructions for agent); Russian optional
• Required YAML frontmatter:

  ---
  name: skill-name
  description: "Description for catalog/search"
  ---
  

• Structure: intro → parameters/steps → examples → constraints/know-how
• No emojis as icons (SVG only)
• Concrete commands, paths, invocation examples

Pre-PR Checklist

# Validate JSON
python3 -m json.tool index.json >/dev/null
python3 -m json.tool skills/<name>/skill.json >/dev/null

# Check required files exist
ls skills/<name>/SKILL.md skills/<name>/skill.json

📄 Project Info