Skip to content

Repository files navigation

my personal ai skills

A personal two-part Skill & agents.md registry:

  • A single-binary Rust CLI with TUI (instinctagents) for installing/removing cataloged skills and agents.md snippets in any project.
  • An agentic internal tool for maintainers that ingests entries from GitHub URLs into this registry repo.

Supports the Claude Code, Codex, OpenCode, and Codebuff harnesses.

Install

curl -fsSL https://raw.githubusercontent.com/uinstinct/aiskills/main/install.sh | bash

This downloads the latest release binary into the current directory as ./instinctagents.

Supported platforms: macOS arm64 (Apple Silicon), macOS x86_64 (Intel), Linux x86_64, Linux aarch64.

debugging for macos installation > **macOS users:** The binary is unsigned, so Gatekeeper may block it on first run. To allow it, either right-click the binary and choose **Open**, or run: > ```sh > xattr -d com.apple.quarantine ./instinctagents > ```

Usage

Run the binary inside any project that uses Claude Code, Codex, OpenCode, or Codebuff:

./instinctagents

The TUI opens on the Add tab. Use Tab / Shift+Tab to switch between Add, Remove, List, and Update.

  • Add - browse the catalog and install one or more skills / agents.md snippets.
  • Remove - uninstall items previously installed by instinctagents.
  • List - read-only view of what's installed.
  • Update - check whether a newer version of the binary is available.

The tool detects your harness automatically by looking for files like CLAUDE.md, AGENTS.md, or opencode.json. Codebuff is matched last, via a .agents/ directory or a knowledge.md file at the project root, so it never shadows a Claude Code, Codex, or OpenCode signal when those coexist.

Repository layout

assets/skills/          # one folder per skill, each with manifest.yml + SKILL.md
assets/agents.md/       # one folder per agents.md integration, each with manifest.yml + snippet.md
src/external/           # Rust source for the user-facing CLI/TUI binary
src/internal/           # Python scripts used by maintainer slash commands
mapping.yml             # registry catalog: every published skill and integration
.github/workflows/      # CI and release workflows

For maintainers

Prerequisites: Install uv — it is the only Python tool you need. uv manages the interpreter and dependencies automatically.

Slash commands shipped in .claude/commands/, .codex/skills/, and .opencode/commands/ cover two flows:

Ingest from a GitHub URL:

/skill-add <github-url>          # fetch a skill from GitHub and add it to the registry
/agents-md-add <github-url>      # fetch an agents.md integration from GitHub and add it to the registry

Scaffold a new local-only entry:

/skill-new <name>                # interactively author a new local skill under assets/skills/<name>/
/agents-md-new <name>            # interactively author a new local agents.md integration under assets/agents.md/<name>/

All four delegate to the Python scripts under src/internal/, which are invoked via uv run --project src/internal.

License

See LICENSE.

About

ai skills used

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages