luongnv89/asm

The universal skill manager for AI coding agents.

What it solves

agent-skill-manager (asm) addresses the fragmentation and overhead of managing "skills" (instruction sets or rules) across multiple AI agents and tools. It eliminates the need to manually copy files into various hidden directories for tools like Claude Code, Cursor, and Codex, while providing a centralized way to to search, install, and audit these skills.

It specifically tackles the "attention budget" problem: since many agents include all installed skill descriptions in their system prompt, asm helps users identify and demote skills that consume too many tokens without providing enough value.

How it works

asm acts as a scriptable CLI and delivery infrastructure. It manages a local library of skills and symlinks or copies them into the specific directories required by 21 different AI agent providers.

Key mechanisms include:

  • Resolution Ladder: When fetching a skill, it searches through installed skills, the local library, an indexed catalog, and a remote registry.
  • Security Scanning: Every installation or fetch performs a pre-install scan for shell execution, network access, and credential exposure.
  • Reference Tier: Users can retrieve a skill's content via asm get to use it once without permanently installing it into an agent's resident context.
  • JSON Interface: Most commands support --json and --yes flags, making the CLI easily consumable by other AI agents or CI/CD pipelines.

Who it’s for

  • AI Agent Users: People using multiple AI coding assistants (e.g., Cursor, Claude Code, Aider) who want a unified way to manage their custom rules and skills.
  • AI Agent Developers: Those building agents that need to dynamically pull in specific capabilities or dependencies just-in-time.
  • Skill Authors: Developers creating SKILL.md files who need tools for scaffolding, linking, and publishing.

Highlights

  • Cross-Provider Support: Manages skills for 21 different agent providers in one place.
  • Attention Budget Tracking: Estimates resident vs. body token costs to help optimize prompt context.
  • Security Audits: Built-in scanning to prevent risky manual installs from GitHub.
  • Audit Tools: Detects redundant skills and semantic overlap (skills doing the same job).
  • Just-in-Time Dependencies: Allows agents to discover and acquire optional dependencies for a specific task session.
  • TUI Dashboard: Includes an optional terminal user interface for local browsing.

Related

  • Project
    dzhng/skills

    A library of reusable, composable AI‑agent “skills” (prompt‑driven actions) for building autonomous software‑factory loops, installable via npm and usable with many coding assistants.

  • Project
    rorkai/app-store-connect-cli-skills

    A collection of agent-ready skills for the App Store Connect CLI that enables AI assistants to automate iOS/macOS release, metadata, and distribution workflows.

  • Project
    MoizIbnYousaf/ai-agent-skills

    A genuine npm‑based CLI/TUI that ships a curated library of AI‑agent skills, lets you build and manage your own skill collections, and integrates with many LLM‑agents via the Agent Skills spec.

  • Project
    vercel-labs/skills

    A CLI that discovers, installs, updates, and runs reusable prompt bundles (“skills”) for many AI coding assistants (Claude Code, Cursor, OpenCode, etc.), pulling them from any Git repository or direct URL.

  • Project
    darkrishabh/agent-skills-eval

    agent‑skills‑eval is a TypeScript/Node.js CLI + SDK that runs an Agent Skill both with and without its `SKILL.md` context, grades the two outputs with a separate LLM judge, and produces JSON artifacts plus a static HTML report. It supports OpenAI‑compatible APIs, custom providers, tool‑call assertions, and full agentskills.io spec validation, making it suitable for skill development, CI testing, and comparative analysis.