Product Requirements Document — agent-skill-manager banner
luongnv89 luongnv89

Product Requirements Document — agent-skill-manager

Development community intermediate

Description

**Product:** agent-skill-manager (`asm`) **Version:** 1.10.0 (current) → 3.0 (vision) **Author:** luongnv89 **Date:** 2026-03-19 **Status:** Active development **License:** MIT ---

Installation

This entry records only its repository, not the path inside it, so there is no exact command to give. Open the source below and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

Repository README

This is the README for luongnv89/asm, shared by 5 entries in this directory. It describes the repository, not this entry specifically.

Product Requirements Document — agent-skill-manager

**Product:** agent-skill-manager (`asm`) **Version:** 1.10.0 (current) → 3.0 (vision) **Author:** luongnv89 **Date:** 2026-03-19 **Status:** Active development **License:** MIT


1. Product Vision

**agent-skill-manager** is the **npm for AI agent skills** — a universal package manager that lets developers install, discover, audit, update, and share reusable skills across every AI coding agent from a single tool.

One-line pitch

One command to manage every AI agent's skills.

North star metric

Become the install command that every skill README links to: `asm install skill-name`.

Target audience

  • Primary: Developers who use 2+ AI coding agents (Claude Code, Codex, Cursor, Windsurf, etc.) and accumulate skills across them.
  • Secondary: Skill authors who want distribution and discoverability for their skills.
  • Tertiary: Teams and organizations that need standardized skill sets across developers.

2. Problem Statement

AI coding agents (Claude Code, Codex, Cursor, Windsurf, OpenClaw, Cline, Roo Code, etc.) each store skills in their own hidden directories with their own conventions. Developers face:

  1. Scattered skills — each agent has its own skill directory (~/.claude/skills/, ~/.codex/skills/, .cursor/skills/, etc.). The same skill ends up installed multiple times in different places.
  2. No visibility — no quick way to see what's installed, what's duplicated, or what's outdated across all agents.
  3. Manual installation — cloning repos, copying folders, hoping SKILL.md is valid, with no security scanning.
  4. No discovery — no central place to find and compare skills. Discovery happens through word-of-mouth and GitHub searches.
  5. No update mechanism — once installed, skills go stale with no notification of available updates.

The problem worsens with every new AI agent adopted.


3. Strategic Position

We are building the **npm analog for AI agent skills** — a lightweight, npm-native package manager that becomes the standard through distribution and ecosystem, not through feature breadth.

**Key advantages we double down on:**

  • Zero-friction npm install (npm i -g agent-skill-manager)
  • Minimal dependencies (2 production deps)
  • Strict TypeScript, 42% test:code ratio
  • Built-in security scanning (26 pattern rules)
  • Pre-indexed offline skill search
  • Lower contribution barrier (TypeScript ecosystem)

4. Existing Features (v1.10.0 — Shipped)

4.1 Dual Interface

Interface Description