@agent-skills-dashboard/core

TypeScript data layer for Agent Skills. It reads SKILL.md files, builds a skill graph, and helps AI agents find or recommend the next skill for a development workflow.

它把 SKILL.md 这类 Agent 技能文件组织成“阶段 + 依赖 + 推荐”的技能图谱。对 AI Agent 来说，它像一个开发流程导航器：可以根据当前阶段、任务描述和技能依赖，推荐下一步应该使用的技能。

Install

pnpm add @agent-skills-dashboard/core

# or
npm install @agent-skills-dashboard/core

AI Install

给 AI Agent 或自动化脚本使用时，可以临时指定淘宝 npm 镜像源，不修改全局 registry：

npm install @agent-skills-dashboard/core --registry=https://registry.npmmirror.com

Quick Start

import {
  createSkillGraph,
  findForPhase,
  findForTask,
  recommendNext,
  createProject,
  updateProjectState,
} from '@agent-skills-dashboard/core';

const graph = await createSkillGraph('/path/to/agent-skills/skills');

const buildSkills = findForPhase(graph, 'build');
const apiSkills = findForTask(graph, 'api integration');
const recommendations = recommendNext(graph, 'define', 'I want to start a new app');

let project = createProject('My AI Project');
project = updateProjectState(project, 'idea-refine', graph);

Input

createSkillGraph() expects a directory that contains skill folders with SKILL.md files:

skills/
├── idea-refine/
│   └── SKILL.md
├── planning-and-task-breakdown/
│   └── SKILL.md
└── shipping-and-launch/
    └── SKILL.md

Main APIs

• createSkillGraph(skillsRoot) parses skills and returns a graph.
• findForPhase(graph, phase) returns skills for define, plan, build, verify, review, or ship.
• findForTask(graph, task) searches skills by keyword, name, and description.
• recommendNext(graph, currentPhase, context) recommends useful next skills.
• createProject(name) creates a project tracking state.
• updateProjectState(project, completedSkillId, graph) marks a skill complete and advances the project phase when ready.

License

MIT