Skip to content

xinetzone/SpecWeave / .agents/skills/aily-pdm

aily-pdm

Flagged

PDM(Python Development Master)专业技能助手,提供完整的Python包管理和项目管理指导。 当用户询问以下内容时使用此技能: - PDM安装、项目初始化、依赖管理(add/remove/update) - pyproject.toml配置、PEP 621元数据 - PDM命令(pdm init, pdm add, pdm install, pdm lock, pdm publish等) - 虚拟环境配置、Python版本管理 - PDM插件开发、脚本编写(tool.pdm.scripts) - 依赖解析错误、构建配置问题 - 从Pipenv/Poetry/requirements.txt迁移到PDM - CI/CD集成、IDE配置(VS Code/PyCharm) - PDM高级用法(hooks、环境变量、集中式缓存) - 项目结构设计、最佳实践 - pdm.lock锁定文件管理 - 发布到PyPI、自定义索引配置 - Python/PDM安装(系统无Python时的引导安装) - conda环境与PDM协同使用 - monorepo多包项目管理 - PEP 723内联脚本、workspace原生支持 - uv后端集成、Docker容器化部署

PDM专业助手

PDM(Python Development Master)是一个现代Python包和依赖项管理器,支持最新PEP标准。

核心能力

  1. 环境准备与安装 - Python/PDM安装引导
  2. 项目初始化与管理
  3. 依赖管理(添加、更新、删除)
  4. 配置与最佳实践
  5. 错误诊断与解决
  6. IDE/CI/CD集成指导
  7. conda环境协同
  8. Monorepo多包管理
  9. PEP 723内联脚本
  10. Workspace原生支持
  11. uv后端集成
  12. Docker容器化部署

第一章:环境准备

1.1 检测当前环境

首先检查系统是否有Python和PDM:

# 检查Python
python --version
python3 --version
which python python3

# 检查PDM
pdm --version
which pdm

# 检查conda
conda --version
which conda

1.2 Python安装

Linux/macOS 安装Python

# 方法1: 使用系统包管理器
# Ubuntu/Debian
sudo apt update && sudo apt install python3 python3-pip python3-venv

# macOS (Homebrew)
brew install python@3.11

# 方法2: 使用pyenv(推荐)
curl https://pyenv.run | bash
# 配置后使用
pyenv install 3.11.8
pyenv global 3.11.8

# 方法3: 使用conda
conda create -n py311 python=3.11 -y
conda activate py311

Windows 安装Python

# 方法1: Microsoft Store(最简单)
# 打开Microsoft Store,搜索"Python 3.11"

# 方法2: 官网下载
# https://www.python.org/downloads/windows/

# 方法3: 使用py launcher(如果安装时勾选了)
py -3.11 --version

# 方法4: winget
winget install Python.Python.3.11

验证安装

python3 --version  # 应显示 Python 3.9+ 或更高版本

1.3 PDM安装

官方推荐安装方式

# Linux/macOS
curl -sSL https://pdm-project.org/install-pdm.py | python3 -

# Windows
powershell -ExecutionPolicy ByPass -c "irm https://pdm-project.org/install-pdm.py | py -"

# 验证安装
pdm --version

其他安装方式

# pipx安装(隔离环境)
pipx install pdm

# pip安装(用户级)
pip install --user pdm

# uv安装(超快速)
uv tool install pdm

# Homebrew (macOS)
brew install pdm

# asdf
asdf plugin add pdm && asdf install pdm latest

安装参数

# 指定版本
curl -sSL https://pdm-project.org/install-pdm.py | python3 - --version 2.10.0

# 安装到指定路径
curl -sSL https://pdm-project.org/install-pdm.py | python3 - --path ~/.local

# 卸载
curl -sSL https://pdm-project.org/install-pdm.py | python3 - --remove

1.4 PDM自升级

pdm self update
pdm self update --head  # 更新到最新开发版

第二章:Conda环境协同

2.1 Conda与PDM的关系

工具职责层级
conda环境管理(Python版本、系统包)底层
PDM包管理(Python包、依赖解析)上层

推荐用法: conda管理Python版本,PDM管理项目依赖

2.2 使用Conda创建的Python环境

# 1. 用conda创建Python环境
conda create -n myproject python=3.11 -y
conda activate myproject

# 2. 在该环境中安装PDM
pip install pdm

# 3. 用PDM初始化项目
pdm init

# 4. 后续用PDM管理依赖
pdm add requests flask

2.3 Conda + PDM 最佳实践

方案A: Conda基础环境 + PDM项目依赖

# 创建基础conda环境
conda create -n base python=3.11 -y
conda activate base

# 安装PDM(用户级)
pip install --user pdm

# 初始化PDM项目
mkdir myproject && cd myproject
pdm init

# PDM管理所有Python包
pdm add numpy pandas scikit-learn

方案B: 每个项目独立Conda环境

# 为项目创建独立conda环境
conda create -n myproject python=3.11 -y
conda activate myproject

# 安装PDM到该环境
pip install pdm

# 初始化项目
pdm init --python 3.11

2.4 PDM识别Conda环境

# PDM会自动检测conda环境
conda activate myenv
pdm info  # 会显示当前使用的Python路径

# 手动指定conda环境的Python
pdm use /home/user/miniconda3/envs/myenv/bin/python

2.5 环境切换脚本

# pyproject.toml
[tool.pdm.scripts]
# 激活conda环境后运行
activate-conda = {shell = "conda activate myproject && pdm run start"}

2.6 常见问题

# Q: conda环境中的PDM无法识别?
# A: 确保在conda激活状态下安装PDM

conda activate myenv
pip install pdm

# Q: 如何让PDM使用conda的包?
# A: PDM不使用conda包,通过conda安装系统依赖(如CUDA)

第三章:项目初始化

3.1 基础初始化

# 交互式初始化
pdm init

# 非交互式初始化(CLI应用)
pdm init --name my-app --python 3.11 --license MIT

# 从模板创建
pdm init --template flask

3.2 项目类型选择

类型特点适用场景
库项目有name/version字段,可发布到PyPI需要分发的包
应用项目无name字段,不可发布内部工具、CLI应用

3.3 Python版本配置

# pyproject.toml
[project]
requires-python = ">=3.9"

注意事项:

  • requires-python 必须覆盖所有依赖项的Python支持范围
  • 依赖项 requires-python 必须包含项目的 requires-python 范围

第四章:依赖管理

4.1 添加依赖

# 添加生产依赖
pdm add requests
pdm add "flask>=2.0"
pdm add requests[socks]  # 带额外依赖

# 添加开发依赖
pdm add -dG test pytest
pdm add -dG lint "flake8; python_version >= '3.9'"

# 添加本地/VCS依赖
pdm add ./my-package
pdm add "git+https://github.com/user/repo.git@main"

4.2 版本保存策略

策略说明示例
--save-minimum最低版本(默认)>=2.21.0
--save-compatible兼容版本>=2.21.0,<3.0.0
--save-exact精确版本==2.21.0
--save-wildcard通配符*

4.3 更新依赖

# 更新所有依赖
pdm update

# 更新指定依赖
pdm update requests

# 忽略版本约束更新
pdm update -u requests

# 更新策略
pdm update --update-eager requests  # 递归更新

4.4 删除依赖

pdm remove requests
pdm remove -G test pytest
pdm remove -dG lint flake8

4.5 依赖覆盖

# pyproject.toml
[tool.pdm.resolution.overrides]
pytz = "2023.3"
requests = ">=2.28.0"

第五章:Monorepo多包项目管理

5.1 Monorepo架构设计

my-monorepo/
├── pyproject.toml          # 根配置
├── pdm.lock               # 统一锁定文件
├── packages/
│   ├── core/              # 核心包
│   │   ├── pyproject.toml
│   │   └── src/
│   ├── api/               # API包
│   │   ├── pyproject.toml
│   │   └── src/
│   └── cli/               # CLI工具包
│       ├── pyproject.toml
│       └── src/
├── scripts/               # 构建脚本
├── docs/                  # 文档
└── tests/                 # 集成测试

5.2 根项目配置

# pyproject.toml (根目录)
[project]
name = "my-monorepo"
version = "0.1.0"
requires-python = ">=3.9"

# 定义所有包的位置
[tool.pdm.resolution]
includes = [
    "packages/core",
    "packages/api",
    "packages/cli",
]

[tool.pdm.scripts]
# 安装所有包(可编辑模式)
install-all = {composite = [
    "pdm install",
    "pdm run --project packages/core install",
    "pdm run --project packages/api install",
    "pdm run --project packages/cli install",
]}

# 清理所有包
clean = {shell = "rm -rf packages/*/__pycache__ packages/*/.venv packages/*/build"}

5.3 子包配置

core/pyproject.toml

# packages/core/pyproject.toml
[project]
name = "@my-monorepo/core"  # 作用域包名
version = "0.1.0"
description = "Core utilities"
requires-python = ">=3.9"
dependencies = []

[tool.pdm]
distribution = true

[build-system]
requires = ["pdm-backend"]
build-backend = "pdm.backend"

api/pyproject.toml

# packages/api/pyproject.toml
[project]
name = "@my-monorepo/api"
version = "0.1.0"
description = "API service"
requires-python = ">=3.9"
dependencies = [
    "@my-monorepo/core>=0.1.0",
    "fastapi>=0.100.0",
    "uvicorn>=0.23.0",
]

[tool.pdm]
distribution = true

cli/pyproject.toml

# packages/cli/pyproject.toml
[project]
name = "@my-monorepo/cli"
version = "0.1.0"
description = "CLI tool"
requires-python = ">=3.9"
dependencies = [
    "@my-monorepo/core>=0.1.0",
    "click>=8.0.0",
]

[project.scripts]
mycli = "my_monorepo_cli:main"

[tool.pdm]
distribution = true

5.4 跨包依赖配置

# packages/api/pyproject.toml
[project]
dependencies = [
    "@my-monorepo/core@ file:///${PROJECT_ROOT}/../core",
    "fastapi>=0.100.0",
]

5.5 Monorepo常用命令

# 根目录:安装所有依赖并锁定
pdm lock

# 根目录:在所有包中搜索
pdm run --project packages/core search requests
pdm run --project packages/api add fastapi

# 根目录:更新所有包
pdm update

# 根目录:构建所有包
for pkg in packages/*/; do
    pdm build --project "$pkg" -d dist/
done

# 发布所有包
pdm publish --project packages/core
pdm publish --project packages/api
pdm publish --project packages/cli

5.6 Monorepo脚本

# pyproject.toml (根目录)
[tool.pdm.scripts]

# 安装所有包为可编辑模式
bootstrap = {shell = """
    pdm install && \
    pdm run --project packages/core install && \
    pdm run --project packages/api install && \
    pdm run --project packages/cli install
"""}

# 发布所有包到PyPI
publish-all = {composite = ["publish-core", "publish-api", "publish-cli"]}

publish-core = "pdm publish --project packages/core"
publish-api = "pdm publish --project packages/api"
publish-cli = "pdm publish --project packages/cli"

# 格式化所有包
format-all = {shell = """
    pdm run --project packages/core run format && \
    pdm run --project packages/api run format && \
    pdm run --project packages/cli run format
"""}

# 运行所有测试
test-all = {composite = ["test-core", "test-api", "test-cli"]}

test-core = "pdm run --project packages/core run test"
test-api = "pdm run --project packages/api run test"
test-cli = "pdm run --project packages/cli run test"

5.7 Monorepo + Git

# .gitignore
__pycache__/
*.py[cod]
*.so
.Python
.venv/
.eggs/
*.egg-info/
dist/
build/
*.egg

# pdm.lock 和 pdm.toml 应该提交
# .pdm-python 不应提交

5.8 依赖组在Monorepo中的应用

# packages/api/pyproject.toml
[dependency-groups]
dev = ["pytest", "pytest-cov", "pytest-asyncio"]
lint = ["ruff", "mypy"]
docs = ["mkdocs", "mkdocs-material"]

# 安装开发依赖
pdm install -G dev -G lint

第六章:PEP 723 内联脚本(PDM 2.16+)

6.1 什么是PEP 723

PEP 723允许在Python脚本顶部声明依赖,运行时自动创建临时环境:

# /// script
# requires-python = ">=3.11"
# dependencies = [
#     "requests",
#     "rich>=13.0",
# ]
# ///
"""
数据分析脚本
直接运行:pdm run script.py
"""
import requests
from rich import print

response = requests.get("https://api.github.com")
print(response.json())

6.2 运行内联脚本

# PDM自动创建临时环境并运行
pdm run script.py

# 指定额外的依赖
pdm run script.py --with numpy,pandas

# 指定Python版本
pdm run script.py --python 3.11

6.3 内联脚本元数据

# /// script
# requires-python = ">=3.10"
# dependencies = ["flask", "sqlalchemy"]
# sources = ["src/", "lib/"]
# ///
字段说明示例
requires-pythonPython版本要求>=3.11
dependencies依赖列表["requests"]
sources额外源码目录["src/"]
dev-dependencies开发依赖["pytest"]

6.4 实际应用场景

# /// script
# requires-python = ">=3.10"
# dependencies = ["httpx", "rich", "pydantic"]
# ///
"""API测试脚本"""
import httpx
from rich.console import Console
from pydantic import BaseModel

console = Console()

class User(BaseModel):
    name: str
    email: str

async def main():
    async with httpx.AsyncClient() as client:
        response = await client.get("https://jsonplaceholder.typicode.com/users/1")
        user = User.model_validate(response.json())
        console.print(f"[green]Name:[/green] {user.name}")
        console.print(f"[blue]Email:[/blue] {user.email}")

import asyncio
asyncio.run(main())

第七章:Workspace原生支持(PDM 2.15+)

7.1 Workspace vs 传统Monorepo

特性传统方式Workspace原生
依赖解析手动配置includes自动发现子包
跨包引用file://路径原生支持
锁定文件手动同步统一管理

7.2 Workspace配置

# pyproject.toml (根目录)
[tool.pdm.workspace]
packages = [
    {path = "packages/core"},
    {path = "packages/api"},
    {path = "packages/cli"},
]

7.3 子包Workspace配置

# packages/api/pyproject.toml
[project]
name = "@my-monorepo/api"
version = "0.1.0"
dependencies = [
    "@my-monorepo/core",  # 直接引用,无需file://
]

[tool.pdm]
is-initializer = false  # 非workspace根

7.4 Workspace命令

# 查看所有包
pdm workspace list

# 查看依赖关系
pdm workspace list --graph

# 构建所有包
pdm build --all

# 发布所有包
pdm publish --all

7.5 Workspace + CI/CD

# .github/workflows/release.yml
- name: Build and publish
  run: |
    pdm install
    pdm workspace list --graph
    pdm publish --all

第八章:uv后端集成(PDM 2.19+ 实验性)

8.1 为什么使用uv

uv是Rust编写的高速包解析器,比传统pip快10-100倍。

8.2 启用uv后端

# 安装uv
pip install uv

# 使用uv作为后端(实验性)
pdm lock --backend uv

# 或在配置中启用
[tool.pdm.resolution]
backend = "uv"

8.3 uv后端限制

# 目前不支持的功能
# 1. VCS依赖需要配置
# 2. 某些特殊索引

[tool.pdm.resolution]
backend = "uv"

8.4 性能对比

操作pipuv
锁定依赖30s1s
安装依赖60s3s
增量更新15s0.5s

第九章:Docker容器化部署

9.1 基础镜像选择

镜像特点适用场景
python:3.11-slim体积小(~150MB)生产环境推荐
python:3.11-alpine最小(~50MB)极致精简
python:3.11完整镜像需要系统依赖

9.2 多阶段构建(推荐)

# Stage 1: 构建阶段
FROM python:3.11-slim as builder

# 安装构建依赖
RUN apt-get update && apt-get install -y --no-install-recommends \
    gcc \
    libffi-dev \
    && rm -rf /var/lib/apt/lists/*

# 安装PDM
RUN pip install pdm

WORKDIR /app

# 复制配置文件
COPY pyproject.toml pdm.lock ./

# 构建依赖(只安装,不复制源码)
RUN pdm install --prod --no-editable --python /app/.venv/bin/python

# Stage 2: 运行阶段
FROM python:3.11-slim as runner

# 只复制虚拟环境和源码
COPY --from=builder /app/.venv /app/.venv
COPY . .

# 设置环境
ENV PATH="/app/.venv/bin:$PATH"
ENV PYTHONUNBUFFERED=1

# 运行应用
CMD ["pdm", "run", "start"]

9.3 非Root用户运行

FROM python:3.11-slim

RUN pip install pdm

WORKDIR /app

# 创建非root用户
RUN useradd --create-home appuser
COPY --chown=appuser:appuser pyproject.toml pdm.lock ./

# 安装依赖
USER appuser
RUN pdm install --prod --python /home/appuser/.local/share/pdm/python/cpython-3.11.5-linux-x86_64-gnu/bin/python3

ENV PATH="/home/appuser/.local/share/pdm/python/cpython-3.11.5-linux-x86_64-gnu/bin:$PATH"
CMD ["pdm", "run", "start"]

9.4 Docker Compose集成

# docker-compose.yml
version: '3.8'

services:
  app:
    build: .
    volumes:
      - .:/app
    environment:
      - PDM_SKIP_SELF=1
    command: pdm run dev

  db:
    image: postgres:15
    environment:
      POSTGRES_PASSWORD: secret

9.5 Docker最佳实践

# 1. 使用特定版本,避免latest
FROM python:3.11.5-slim

# 2. 利用构建缓存
COPY requirements.txt .
RUN pip install -r requirements.txt
COPY . .

# 3. 使用.pdm-python固定Python版本
COPY .pdm-python /app/.pdm-python

# 4. 清理缓存
RUN apt-get clean && rm -rf /var/lib/apt/lists/*

第十章:平台特定锁定

10.1 多平台配置

# pyproject.toml
[tool.pdm.resolution]
targets = [
    "x86_64-linux-gnu",
    "aarch64-linux-gnu",
    "x86_64-apple-darwin",
    "x86_64-pc-windows-msvc",
]

10.2 条件依赖

[project]
dependencies = [
    "numpy>=1.24; platform_system == 'Windows'",
    "pyobjc-framework-Cocoa>=9.0; platform_system == 'Darwin'",
]

第十一章:安全扫描

11.1 依赖漏洞检查

# 安装安全工具
pdm add -dG security pip-audit safety

# 运行扫描
pdm run pip-audit
pdm run safety check

11.2 GitHub Actions集成

- name: Security audit
  run: |
    pip install pip-audit
    pip-audit --strict

第十二章:配置文件结构

12.1 pyproject.toml 基础结构

[project]
name = "my-package"
version = "0.1.0"
description = "A short description"
requires-python = ">=3.9"
dependencies = [
    "requests>=2.28.0",
    "click>=8.0.0",
]

[project.optional-dependencies]
dev = ["pytest>=7.0.0", "black"]
doc = ["mkdocs"]

[dependency-groups]
lint = ["flake8", "mypy"]

[build-system]
requires = ["pdm-backend"]
build-backend = "pdm.backend"

[tool.pdm.distribution]
package = true

[tool.pdm.scripts]
start = "flask run"
test = "pytest"

12.2 pdm.toml 配置

[python]
path = ".venv/bin/python"

[mirror]
url = "https://pypi.tuna.tsinghua.edu.cn/simple"

第十三章:PDM脚本

13.1 脚本类型

[tool.pdm.scripts]
# 命令脚本
start = {cmd = "flask run -p 5000"}

# Shell脚本(支持管道)
filter = {shell = "cat error.log | grep ERROR"}

# Python函数调用
main = {call = "my_package:main"}

# 组合脚本
dev = {composite = ["start", "test"]}

13.2 脚本选项

[tool.pdm.scripts]
start.cmd = "flask run"
start.env = {FLASK_ENV = "development"}
start.env_file = ".env"
start.working_dir = "src"

13.3 钩子脚本

[tool.pdm.scripts]
pre_install = "echo 'Installing...'"
post_install = "echo 'Done!'"
pre_build = "echo 'Building...'"
post_publish = "echo 'Published!'"

第十四章:常用命令速查

命令说明
pdm init初始化项目
pdm add <pkg>添加依赖
pdm remove <pkg>删除依赖
pdm update更新依赖
pdm install安装依赖
pdm lock生成/更新锁定文件
pdm sync同步工作集
pdm run <script>运行脚本
pdm build构建分发包
pdm publish发布到PyPI
pdm list列出依赖
pdm outdated检查过时依赖
pdm info显示项目信息
pdm python install <ver>安装Python版本
pdm self update更新PDM
pdm workspace list列出workspace包
pdm run script.py运行PEP 723脚本

第十五章:虚拟环境管理

# 创建虚拟环境
pdm venv create

# 列出虚拟环境
pdm venv list

# 激活虚拟环境
eval $(pdm venv activate)

# 删除虚拟环境
pdm venv remove

第十六章:从其他工具迁移

16.1 从requirements.txt迁移

pdm import requirements.txt

16.2 从Pipfile迁移

pdm import Pipfile

16.3 从Poetry迁移

Poetry的pyproject.toml部分可直接被PDM读取。


第十七章:CI/CD集成

17.1 GitHub Actions

- uses: pdm-project/setup-pdm@v4
  with:
    python-version: '3.11'
- name: Install dependencies
  run: pdm install
- name: Run tests
  run: pdm run test

17.2 发布配置

- name: Publish to PyPI
  run: pdm publish
  env:
    PDM_PUBLISH_REPO: pypi
    PDM_PUBLISH_USERNAME: __token__
    PDM_PUBLISH_PASSWORD: ${{ secrets.PYPI_API_TOKEN }}

17.3 GitLab CI

# .gitlab-ci.yml
image: python:3.11-slim

before_script:
  - pip install pdm
  - pdm use 3.11

test:
  script:
    - pdm install
    - pdm run test

publish:
  script:
    - pdm build
    - pdm publish
  only:
    - tags

第十八章:常见问题诊断

18.1 Resolution Impossible

Resolution Impossible: The project's requires-python doesn't allow all dependencies

解决:调整requires-python范围,确保覆盖所有依赖项的Python要求。

18.2 包版本冲突

解决:使用依赖覆盖

[tool.pdm.resolution.overrides]
problematic-package = ">=1.0.0"

18.3 构建后端问题

确保build-system配置正确:

[build-system]
requires = ["pdm-backend>=2.0.0"]
build-backend = "pdm.backend"

18.4 Python/PDM未安装

解决:参考第一章环境准备部分进行安装

18.5 Conda环境问题

# Q: PDM无法找到conda的Python
# A: 确保conda环境已激活

conda activate myenv
pdm use python  # 让PDM检测当前Python

# 或手动指定
pdm use /path/to/conda/envs/myenv/bin/python

18.6 uv后端兼容性问题

# Q: 使用uv后端后某些依赖安装失败
# A: 降级到标准后端

pdm lock --backend pdm

第十九章:最佳实践

  1. 始终提交pdm.lock - 确保团队使用相同依赖版本
  2. 合理设置requires-python - 避免不必要的版本限制
  3. 使用依赖组 - 分离生产/开发依赖
  4. 启用脚本钩子 - 自动化重复任务
  5. 配置包索引 - 企业使用私有源提升速度
  6. Monorepo结构清晰 - 使用workspace原生支持
  7. conda + PDM协同 - conda管理Python环境,PDM管理依赖
  8. PEP 723内联脚本 - 快速原型开发和数据分析
  9. Docker多阶段构建 - 生产镜像更小更安全
  10. 定期安全扫描 - 使用pip-audit检查漏洞
  11. 使用uv后端 - 加速CI/CD构建

参考文档

writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 321kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 197kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 118kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall