apm: A Package Manager for AI Agents

# macOS/Linux
curl -sSL https://aka.ms/apm-unix | sh
# Windows (PowerShell)
irm https://aka.ms/apm-windows | iex

$ apm --version
 
Agent Package Manager (APM) CLI version 0.8.11 (81082e2)

apm install microsoft/apm-sample-package

apm install microsoft/apm-sample-package#v1.0.0

name: apm-test
version: 1.0.0
description: APM project for apm-test
author: azukiazusa1
dependencies:
  apm:
  - microsoft/apm-sample-package
  mcp: []
scripts: {}

lockfile_version: '1'
generated_at: '2026-04-16T11:10:34.684248+00:00'
apm_version: 0.8.11
dependencies:
- repo_url: microsoft/apm-sample-package
  host: github.com
  resolved_commit: fb2851683be0e0e7711421d518bd8dba23b0b1f6
  package_type: apm_package
  deployed_files:
  - .github/agents/design-reviewer.agent.md
  - .github/instructions/design-standards.instructions.md
  - .github/prompts/accessibility-audit.prompt.md
  - .github/prompts/design-review.prompt.md
  - .github/skills/style-checker
  content_hash: sha256:744cca54cc8ff7ca90aa1dd621c2f98c6291cd793815afe8518001cc94b8aba9
- repo_url: github/awesome-copilot
  host: github.com
  resolved_commit: 9d117370f901e2244316dfd8cf5362435705b658
  virtual_path: skills/review-and-refactor
  is_virtual: true
  depth: 2
  resolved_by: microsoft/apm-sample-package
  package_type: claude_skill
  deployed_files:
  - .github/skills/review-and-refactor
  content_hash: sha256:9236d06a1500089ddb46975b866e9a63478e502afe7095b1980c618678a7c7fe

apm_modules
├── github
│   └── awesome-copilot
│       └── skills
│           └── review-and-refactor
│               └── SKILL.md
└── microsoft
    └── apm-sample-package
        ├── apm.yml
        ├── CODE_OF_CONDUCT.md
        ├── CONTRIBUTING.md
        ├── LICENSE
        ├── README.md
        └── SECURITY.md
 
7 directories, 7 files

.github
├── agents
│   └── design-reviewer.agent.md
├── instructions
│   └── design-standards.instructions.md
├── prompts
│   ├── accessibility-audit.prompt.md
│   └── design-review.prompt.md
└── skills
    ├── review-and-refactor
    │   └── SKILL.md
    └── style-checker
        └── SKILL.md
 
7 directories, 6 files

apm install microsoft/apm-sample-package --target claude

.claude
├── agents
│   └── design-reviewer.md
├── commands
│   ├── accessibility-audit.md
│   └── design-review.md
├── rules
│   └── design-standards.md
└── skills
    ├── review-and-refactor
    │   └── SKILL.md
    └── style-checker
        └── SKILL.md
 
7 directories, 6 files

scripts:
  style-check: 'claude "/style-checker"'

apm run style-check

apm init writing-assistant

name: writing-assistant
version: 1.0.0
description: APM project for writing-assistant
author: azukiazusa1
dependencies:
  apm: []
  mcp: []
scripts: {}

.apm
├── agents
│   └── *.agent.md
├── instructions
│   └── *.instructions.md
├── prompts
│   └── *.prompt.md
├── skills
│   └── <skill-name>
│       └── SKILL.md
└── hooks
    ├── *.json
    └── scripts
        └── *.sh, *.py ...

---
description: "Instructions for writing technical articles. Defines style and structure guidelines for writing technical articles."
applyTo: "articles/*.md"
---
 
### Article structure
 
1. **Introduction** — Start with "This article..." and briefly describe the theme, background, and purpose
2. **Body** — Develop logically using `##` / `###` headings. Each section explains specific technical elements or procedures
3. **Summary** — Reconfirm key points in bullet points. Use the heading `## Summary`
4. **References** — List referenced official documents and articles under the heading `## References`
 
### Tone
 
- Use a professional yet accessible tone
- Prioritize objective explanations. Use phrases like "I think..." or "I felt..." for personal opinions
- Target audience is engineers interested in web development. Assume some foundational knowledge but explain new concepts thoroughly
- Use transition words like "however," "also," "for example," "therefore," and "next" to make the logical flow clear
 
### Markdown conventions
 
**Code blocks**
 
Always include a language identifier. To indicate a file name, append `:filename`.
 
```ts:src/index.ts
// code
```
 
```
 
To show diffs, use `diff`.
 
```
 
```diff
- deleted line
+ added line
```
 
**Callouts**
 
```
:::info
Supplementary information
:::
 
:::tip
Tips and recommendations
:::
 
:::warning
Warnings and cautions
:::
```

---
name: tech-review
description: A skill for reviewing technical books and manuscripts for software engineers, providing improvement suggestions from the perspectives of structure, technical accuracy, and reader experience. Triggered when users request reviews of manuscripts, chapter drafts, tutorial articles, or technical documents. Handles all review work related to technical book production, including writing support, editorial feedback, structural improvement, code review (code within books), and reader level adjustment.
---
 
# Technical Book Manuscript Review Skill
 
Review manuscripts for technical books, tutorials, and technical articles for software engineers from a professional editor's perspective, providing specific improvement suggestions.
 
## Overall Review Flow
 
Upon receiving a manuscript, proceed with the review in the following order.
 
### 1. Initial Analysis (What to Understand First)
 
Before reading the manuscript, confirm or estimate the following:
 
- **Target readers**: Beginner / Intermediate / Advanced level
- **Book type**: Introductory / Reference / Cookbook / Tutorial / Practical guide
- **Target technology**: Language, framework, tools, etc.
- **Manuscript stage**: First draft / Second draft / Near final
 
If the user hasn't specified these, estimate from the manuscript content and confirm: "I'll review this as an intermediate-level tutorial on XX — does that match your intent?"
 
### 2. Macro Review (Structure and Big Picture)
 
Start with the big picture. Save detailed grammar and typos for later.
 
**Perspectives for checking structural logic:**
 
- Is there a logical flow in the arrangement of chapters and sections? Can readers naturally understand "why this order?"
- Does the content build progressively, where concepts learned in earlier chapters are applied in later ones?
- Do difficult concepts appear suddenly? (Sharp difficulty jumps)
- Is the volume balance between chapters appropriate? (No extremely long or short chapters)
 
**Scope appropriateness:**
 
- Is the content too broad or too narrow for the book's title and purpose?
- Is it clear "what the reader will be able to do after reading this book?"
 
### 3. Micro Review (Quality of Each Chapter/Section)
 
#### Technical Accuracy
 
This is the most important aspect. It directly impacts the credibility of the technical book.
 
- Are there syntax or logic errors in code samples?
- Are API and library versions specified? Are deprecated features being used?
- Are there errors or inaccurate descriptions in technical explanations?
- Are security anti-patterns being presented as best practices?
 
Verify whether code actually works to the extent possible. If code can be run in a container environment, try running it.
 
#### Reader Experience
 
- When introducing new concepts, is the motivation "why this is needed" presented first?
- Are definitions or explanations provided when specialized terms first appear?
- Do concrete examples follow abstract explanations?
- Are there signposts like "summary so far" or "what to learn next?"
 
#### Code Sample Quality
 
Code in technical books isn't sufficient just because it "works." Since it's meant for readers to learn:
 
- Do variable and function names express intent? (Avoid meaningless names like `x`, `tmp`)
- Is the structure progressively more complex? (Start simple, gradually add features)
- Do comments explain "why" rather than "what?"
- Is error handling included? (Instill production code habits)
- Can it be copied and pasted and run? (No missing imports, undefined variables)
 
### 4. Consistency Check
 
Verify uniformity across the entire book:
 
- **Terminology consistency**: Are different terms used for the same concept? (e.g., mixing "function" and "method," mixing "deploy" and "deployment")
- **Code style consistency**: Are indentation, naming conventions, and language versions consistent across chapters?
- **Writing style consistency**: Is there mixing of formal and informal styles?
- **Figure/table formatting**: Are captions, numbering, and reference methods unified?
 
### 5. Review Report Output
 
Output the review results in the following structure:
 
```
# Manuscript Review Report
 
## Overview
- Target: [Manuscript title/chapter number]
- Target readers: [Level]
- Overall assessment: [Brief overall evaluation]
 
## Strengths
[Specifically highlight the manuscript's strengths. Sharing what works well maintains the author's motivation]
 
## Critical Improvement Suggestions (Priority: High)
[Technical errors, major structural issues — must fix before publication]
 
## Improvement Suggestions (Priority: Medium)
[Reader experience improvements, clarification — would make it better]
 
## Minor Notes (Priority: Low)
[Typos, expression tweaks, preference-level suggestions]
 
## Code Review
[Specific feedback for each code sample. Include corrected code]
 
## Next Steps
[Actions the author should take next]
```
 
## Review Approach
 
- **Respect the author**: The manuscript is a culmination of the author's expertise and effort. Aim for constructive feedback, not criticism
- **Be the reader's advocate**: Always consider "how would someone reading this explanation for the first time feel?"
- **Be specific in suggestions**: Instead of "unclear," say "adding an example of XX would make this easier to understand"
- **Prioritize**: Pointing out everything at the same importance level overwhelms the author. Start with critical issues

dependencies:
  mcp:
  - io.github.upstash/context7

apm install

{
  "servers": {
    "io.github.upstash/context7": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@upstash/context7-mcp"
      ],
      "env": {
        "CONTEXT7_API_KEY": "${input:context7-api-key}"
      }
    }
  },
  "inputs": [
    {
      "type": "promptString",
      "id": "context7-api-key",
      "description": "API key for authentication",
      "password": true
    }
  ]
}

git init
git add .
git commit -m "Initial commit"
git remote add origin https://github.com/<username>/writing-assistant
git push -u origin main

apm install azukiazusa1/writing-assistant --target claude

.
├── .claude
│   ├── rules
│   │   └── technical-article.md
│   └── skills
│       └── tech-reviewer
│           └── SKILL.md
├── .gitignore
├── .vscode
│   └── mcp.json
├── apm.yml
└── apm.lock.yaml

---
paths:
- "articles/*.md"
---
 
### Article structure
 
...