Skip to content

Latest commit

 

History

15 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Curated Agent Skills

skills.sh License: MIT

A curated collection of production-grade, harness-agnostic AI agent skills. Built for real engineering workflows with a strict focus on security, token economy, and universal compatibility across all AI coding assistants (Pi, Claude Code, Cursor, Codex, OpenCode, Windsurf, and custom harnesses).


Skills Catalog

Skill Description Runtime Prerequisites
web-search High-signal web search, page extraction, and deep reasoning via Exa with ~80% token compaction and automatic tracking deduplication. Node.js 18+ (ESM) None (EXA_API_KEY optional for Deep API)
secure-url-fetcher Security-hardened HTTP/HTTPS fetcher with pre-resolution IP blacklisting and socket pinning to eliminate SSRF and DNS rebinding risks. Python (uv) uv

Universal Installation

Option 1: Interactive Agent Installer (skills.sh)

Install the complete suite or specific skills into any agent harness (npx, pnpx, or bunx):

# Install all skills
npx skills add r1cc4rd0m4zz4/myskills

# Or install individual skills
npx skills add r1cc4rd0m4zz4/myskills/skills/web-search
npx skills add r1cc4rd0m4zz4/myskills/skills/secure-url-fetcher

Option 2: Pi Coding Agent

Install directly via Pi's native package manager:

pi install git:github.com/r1cc4rd0m4zz4/myskills

Option 3: Local Symlink for Development

Clone the repository and link all skills into your local agent directories (~/.agents/skills/):

git clone https://github.com/r1cc4rd0m4zz4/myskills.git
cd myskills
./scripts/link-skills.sh

Design Principles

  1. Zero Vendor Lock-In: Compliant with the open Agent Skills standard. Every skill is defined in a standard SKILL.md with YAML frontmatter.
  2. Deterministic Token Economy: Raw JSON/HTML payloads are pruned and compressed (~80% context savings) before reaching the LLM context window.
  3. Absolute Privacy: Zero host or environment data leakage; all paths, examples, and fixtures use generic placeholders.
  4. Defense in Depth: Zero unvalidated string interpolation in shell calls; strictly validated network I/O with IP pinning.
  5. Self-Contained Execution: No bloat. Node tools run as zero-dependency standalone .mjs modules; Python tools use inline PEP 723 metadata with uv.

Contributing

See CONTRIBUTING.md for contribution guidelines, and AGENTS.md / CONTEXT.md for architectural and domain standards.


Author & Contributors


License

MIT © Riccardo Mazza

About

Curated, harness-agnostic AI agent skills focused on security, token optimization, and deep web intelligence.

Topics

Resources

Contributing

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages