# PinkSkills
A collection of skills that can be installed directly in AI coding agents. Each skill is distributed in its own directory, with its core behavior defined in the corresponding `SKILL.md`.
[Visit the PinkSkills website →](https://3xian.github.io/PinkSkills/)
## Skills
| Skill | Description | Output language | Invocation |
| --- | --- | --- | --- |
| [`parse-requirements`](./parse-requirements/) | Breaks down PRDs, requirement documents, screenshots, or wireframes into structured implementation requirements, then maps them to modules, implementation approaches, and verification steps in the current codebase | English | `/parse-requirements` |
| [`parse-requirements-cn`](./parse-requirements-cn/) | Chinese version of `parse-requirements` with the same evidence-driven workflow and a Chinese output template | Chinese | `/parse-requirements-cn` |
| [`pink-review`](./pink-review/) | Automatically determines the Git change scope and performs evidence-based defect review plus behavior-preserving simplification review | English | `/pink-review` |
| [`pink-review-cn`](./pink-review-cn/) | Chinese version of `pink-review` that reports review findings in Chinese | Chinese | `/pink-review-cn` |
## Installation
Choose a skill and copy its entire directory into a skills directory the host agent scans. Preserve the directory structure. For example:
```text
<skills-directory>/
└── pink-review/
├── SKILL.md
├── README.md
└── VERSION
```
Typical locations:
- Project: `<project>/.agents/skills/` or `<project>/.<agent>/skills/`
- User: `~/.agents/skills/` or `~/.<agent>/skills/`
`<agent>` is the host agent name. On Windows, `~` is `%USERPROFILE%`. After installation, invoke the skill with the command shown in the table above.
## Usage
### Parse requirements
Run `/parse-requirements` for English output or `/parse-requirements-cn` for Chinese output, then provide a path to a requirement document, paste the requirement text, or attach a file in chat. Both skills will:
1. Read all text and images, prioritizing boxed, circled, highlighted, and arrow-marked content.
2. Break the document into atomic requirements that can be implemented and verified independently.
3. Inspect the current codebase to identify the actual modules, current behavior, and implementation constraints.
4. Write the parsed result to `.requirements/<source-stem>-parsed.md` by default.
Each requirement includes a one-sentence summary, the current state and rationale, affected modules, implementation steps, and verification steps.
### Review code
Run `/pink-review` or `/pink-review-cn`. No arguments are required. The skill will automatically:
1. Determine the current branch, comparison base, and committed, staged, unstaged, and untracked changes.
2. Review correctness, security, compatibility, error handling, performance, and maintainability risks.
3. Run a separate simplification pass to find opportunities to remove branches, state, duplicated flows, or unnecessary abstractions.
4. Run the smallest relevant checks supported by the environment and explicitly disclose any coverage gaps.
The code review skills are always read-only. They never modify code or create commits.
## Repository structure
```text
.
├── .github/
│ └── workflows/
│ └── pages.yml
├── parse-requirements/
│ └── SKILL.md
├── parse-requirements-cn/
│ └── SKILL.md
├── pink-review/
│ ├── SKILL.md
│ ├── README.md
│ └── VERSION
├── pink-review-cn/
│ ├── SKILL.md
│ ├── README.md
│ └── VERSION
├── site/
│ ├── index.html
│ ├── script.js
│ └── styles.css
└── README.md
```
Each directory contains an independent skill. When modifying a skill, update that directory's documentation and version information, if present, at the same time.