# Vllm Rfc Generator

> Generate a vLLM-style RFC (Request for Comments) document based on user input. Use when the user wants to create an RFC for a major architectural change or design decision in vLLM. Triggered by requests like "帮我生成一个vLLM RFC", "create a vLLM RFC for ...", "写一个RFC关于...", "生成RFC文档".

- **Type:** Skill
- **Install:** `agentstack add skill-shen-shanshan-vllm-dev-skills-vllm-rfc-generator`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [shen-shanshan](https://agentstack.voostack.com/s/shen-shanshan)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** Apache-2.0
- **Upstream author:** [shen-shanshan](https://github.com/shen-shanshan)
- **Source:** https://github.com/shen-shanshan/vllm-dev-skills/tree/master/skills/vllm-rfc-generator
- **Website:** https://zhuanlan.zhihu.com/p/2031696581678866733

## Install

```sh
agentstack add skill-shen-shanshan-vllm-dev-skills-vllm-rfc-generator
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# vLLM RFC Generator

Generate a vLLM-style RFC Markdown document from user-provided information and save it to `./outputs/`.

## Inputs

Collect the following from the user (ask if not provided):

| Field | Required | Description |
|-------|----------|-------------|
| **RFC Title** | Yes | Short, descriptive title (will be prefixed with `[RFC]:`) |
| **Main Content** | Yes | Description of the motivation and proposed changes |
| **Related PRs** | No | List of PR numbers or URLs; classify them by category |
| **Feedback Period** | No | Duration for feedback (default: "At least 1 week") |
| **CC List** | No | GitHub usernames to CC (e.g., `@user1`, `@user2`) |
| **Other Notes** | No | Any additional context or notes |

## Workflow

1. **Collect inputs** — gather all necessary information from the user
2. **Classify PRs** — if PRs are provided, group them by category (see PR Classification below)
3. **Generate RFC content** — write each section following the template
4. **Save output** — write to `./outputs/rfc-.md`
5. **Confirm** — tell the user the output path

## PR Classification

When the user provides related PRs, classify them into appropriate categories such as:

- **Background / Motivation PRs** — existing issues, bugs, or prior discussions that motivate this RFC
- **Implementation PRs** — concrete code changes implementing the proposed design
- **Dependency PRs** — prerequisite changes that must land first
- **Related / Follow-up PRs** — tangential improvements or future work inspired by this RFC

Format each category as a bullet list:

```markdown
### Related PRs

**Background / Motivation:**
- [#XXXX](https://github.com/vllm-project/vllm/pull/XXXX): Brief description

**Implementation:**
- [#YYYY](https://github.com/vllm-project/vllm/pull/YYYY): Brief description

**Dependencies:**
- [#ZZZZ](https://github.com/vllm-project/vllm/pull/ZZZZ): Brief description
```

If a PR number has no description, fetch its title via the GitHub API or `gh` CLI:
```bash
gh pr view  --repo vllm-project/vllm --json title,url -q '"\(.url): \(.title)"'
```

## RFC Document Template

Save to `/Users/shanshan-shen/.claude/skills/vllm-rfc-generator/outputs/rfc-.md`
where `` is the title converted to lowercase with hyphens (e.g., `rfc-async-engine-redesign.md`).

```markdown
# [RFC]: 

> **Labels**: RFC
> **Status**: Open for Feedback
> **Feedback Period**: 

---

## Motivation

- What problem does it solve?
- What are the current limitations or pain points?
- Why is this the right time to address it?

## Proposed Change

Key points to cover:
- What is being changed / added / removed?
- High-level design or architecture
- API surface changes (if any)
- Migration path or backward compatibility

### Key Design Decisions

- **Decision 1**: Rationale
- **Decision 2**: Rationale

## Related PRs

## Feedback Period

## CC List

, 

## Any Other Things

---

> Please take a look at previous [RFCs](https://github.com/vllm-project/vllm/issues?q=label%3ARFC+sort%3Aupdated-desc) for reference.
```

## Writing Guidelines

- **Motivation**: Be concrete. Reference real pain points, GitHub issues, or user feedback. 2–4 short paragraphs.
- **Proposed Change**: Be specific enough that reviewers can evaluate the design, but avoid implementation minutiae. Use subsections for complex proposals.
- **Tone**: Professional, neutral, open to feedback. This is a proposal, not a decree.
- **Length**: Aim for 400–800 words total across Motivation and Proposed Change. Quality over quantity.
- **Language**: Write in **English** by default. If the user requests Chinese, write in Chinese.

## Example Output Path

```
/Users/shanshan-shen/.claude/skills/vllm-rfc-generator/outputs/rfc-async-engine-redesign.md
```

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [shen-shanshan](https://github.com/shen-shanshan)
- **Source:** [shen-shanshan/vllm-dev-skills](https://github.com/shen-shanshan/vllm-dev-skills)
- **License:** Apache-2.0
- **Homepage:** https://zhuanlan.zhihu.com/p/2031696581678866733

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-shen-shanshan-vllm-dev-skills-vllm-rfc-generator
- Seller: https://agentstack.voostack.com/s/shen-shanshan
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.