Open Skills

Battle-tested execution playbooks that give any AI agent the exact commands, APIs, and patterns it needs — cutting token usage by 95–98% and making local models as capable as GPT-4.

---

The Problem

AI agents waste tokens discovering what you already know:

• Cloud models (GPT-4, Claude) — 10–30 trial-and-error calls per task → $0.15–$0.25/task
• Local models (Llama, Mistral, Qwen) — often fail outright without step-by-step guidance

The Solution

Pre-written, tested skill files your agent reads once and executes correctly the first time.

Without Open Skills                 With Open Skills
─────────────────────────────────   ─────────────────────────────────
Agent searches for API docs         Agent reads SKILL.md
Tries wrong endpoint                Runs the exact working command
Debugs response format              Parses the output correctly
Retries 15–20 times                 Done in 1–3 calls

~50,000 tokens  ~$0.20              ~1,000 tokens  ~$0.004

---

Quick Start

Step 1 — Clone the skills library (once)

git clone https://github.com/besoeasy/open-skills ~/open-skills

Step 2 — Tell your agent about it

Copy the contents of prompt.txt into your agent's system prompt, memory, or instructions file. This tells the agent to check ~/open-skills before every task and auto-sync skills from the repo.

---

OpenCode

Drop prompt.txt into your project's AGENTS.md file. OpenCode reads it automatically at session start — zero extra configuration.

cat prompt.txt >> AGENTS.md

---

Claude Desktop

1. Open Claude Desktop → Settings → Custom Instructions
2. Paste the full contents of prompt.txt
3. Save — Claude will check ~/open-skills before every task in every conversation

---

Cursor

Add the contents of prompt.txt to your global user rules:

1. Open Cursor → Settings → General → Rules for AI
2. Paste the contents of prompt.txt

Or add it per-project via .cursorrules:

cat prompt.txt >> .cursorrules

---

Windsurf

Add the contents of prompt.txt to your global user rules:

1. Open Windsurf → Settings → AI → Custom Instructions
2. Paste the contents of prompt.txt

Or add it per-project via .windsurfrules:

cat prompt.txt >> .windsurfrules

---

GitHub Copilot (VS Code)

Add a .github/copilot-instructions.md file to your repository:

cp prompt.txt .github/copilot-instructions.md

Copilot reads this file as context for every chat in that workspace.

---

Aider

Pass prompt.txt as a system prompt when launching:

aider --system-prompt "$(cat ~/open-skills/prompt.txt)"

Or add it to your ~/.aider.conf.yml:

system-prompt: /home/you/open-skills/prompt.txt

---

Continue.dev

Add an entry to your ~/.continue/config.json under systemMessage:

{
  "models": [...],
  "systemMessage": "<paste contents of prompt.txt here>"
}

---

Any other agent

Paste the full contents of prompt.txt as the system prompt. The instructions are agent-agnostic — they work with any model that can read files and run shell commands.

---

Works best with OpenCode.ai — drop prompt.txt into your project's AGENTS.md and the agent picks up every skill automatically, with zero extra configuration.

---

Skills

Skill	What it does
age-file-encryption	Encrypt / decrypt files with age
anonymous-file-upload	Upload files without an account
browser-automation-agent	Automate browsers with Playwright
bulk-github-star	Star GitHub repos in bulk via CLI
changelog-generator	Generate changelogs from git history
chat-logger	Log and persist chat conversations
check-crypto-address-balance	Look up Bitcoin / crypto balances
city-distance	Calculate distance between cities
city-tourism-website-builder	Build a tourism site for any city
csv-data-summarizer	Summarize and analyze CSV files
d3js-data-visualization	Create charts with D3.js
database-query-and-export	Query databases and export results
file-tracker	Track file changes over time
free-geocoding-and-maps	Geocode addresses for free
free-translation-api	Translate text without API keys
free-weather-data	Get weather data for free
generate-asset-price-chart	Chart stock / crypto price history
generate-qr-code-natively	Generate QR codes with no service
get-crypto-price	Fetch live crypto prices
humanizer	Make AI-written text sound human
ip-lookup	Look up IP address geolocation
json-and-csv-data-transformation	Transform between JSON and CSV
news-aggregation	Aggregate news from RSS / APIs
nostr-logging-system	Log events to the Nostr network
pdf-manipulation	Merge, split, and edit PDFs
phone-specs-scraper	Scrape phone specs from the web
presenton	Generate presentations from text
random-contributor	Pick a random repo contributor
send-email-programmatically	Send email from a script
static-assets-hosting	Host static files for free
trading-indicators-from-price-data	Calculate RSI, MACD, and more
torrent-search	Find torrents by title or IMDB ID
user-ask-for-report	Generate structured reports on demand
using-nostr	Read and post on Nostr
using-scrapy	Scrape websites with Scrapy
using-telegram-bot	Build and run Telegram bots
using-web-scraping	General web scraping patterns
using-youtube-download	Download YouTube videos / audio
web-interface-guidelines-review	Review UI against best practices
web-search-api	Search the web free via SearXNG

---

Cost Impact

Setup	Cost / task	Success rate	Privacy
Cloud model, no skills	$0.15 – $0.25	85 – 95%	❌ Cloud
Cloud model + Open Skills	$0.003 – $0.005	~98%	❌ Cloud
Local model, no skills	$0	30 – 50%	✅ Local
Local model + Open Skills	$0	~95%	✅ Local

The 100% free stack:

curl -fsSL https://ollama.com/install.sh | sh
ollama pull llama3.1:8b
git clone https://github.com/besoeasy/open-skills ~/open-skills
# GPT-4-level task execution — $0 cost, fully offline

---

Why It Works

Skills separate reasoning from execution knowledge:

• The model handles intent and orchestration
• Open Skills provides the tested commands, API patterns, and parsing logic
• Result: fewer retries, lower token usage, higher reliability

Every skill file is:

• ✅ Production-tested — real working code, not theory
• ✅ Agent-optimized — structured for direct LLM consumption
• ✅ Privacy-first — free public APIs, no vendor lock-in
• ✅ Model-agnostic — works with GPT-4, Claude, Llama, Mistral, Gemini, anything

---

Contributing

See CONTRIBUTING.md and SKILL_TEMPLATE.md.

Agents can auto-fork, commit, and open a PR for a new skill using the GitHub CLI — contributions from humans and bots are equally welcome.

---

MIT License