dockershrink-requirements

Find unused packages in requirements.txt that are not needed by the project code

安装方式

默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

直接命令不会经过审查 Prompt；运行前请先检查来源。

npx skills add https://github.com/causify-ai/helpers --skill dockershrink-requirements

命令会保持在同一行。复制前请横向滚动并检查完整内容。

想先保存到本地？可下载 SkillsMP 当前能够提供的文件。

决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。

SKILL.md

来源说明 · 只读预览

Refactor unit test files by aligning strings, renaming methods, and factoring out common test code

Identify and refactor duplicated code blocks into shared functions across Python files

Find documentation files for a given dir, file, class, or function and summarize in 3 bullet points

name	docker.shrink_requirements
description	Find unused packages in requirements.txt that are not needed by the project code

You are an expert of Docker
I will pass you a directory $TARGET with a project
Find out which packages in requirements.txt are not needed by the code in $TARGET

Use grep -r "import <pkg>" or AST-based analysis to find actual imports across all .py files in $TARGET
Map package names in requirements.txt to their import names (e.g., Pillow → PIL, scikit-learn → sklearn, PyYAML → yaml)
Also check setup.py, pyproject.toml, tasks.py, and Makefile for indirect or tool-level usage
Flag but do NOT remove packages that are:
- Runtime plugins or extras loaded dynamically (e.g., via importlib)
- Transitive dependencies pulled in by other listed packages
- Used only in test files (mark them as "test-only")

Print a table with columns:
```
package | import_name | used_in_code | verdict
```
- verdict is one of: remove, keep, investigate, test-only
Propose the trimmed requirements.txt contents and write it to disk