Token导航 LogoToken导航

Agent Skill · 研究检索

feature-vision(特征视觉)

feature-vision 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

feature-vision 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为需要联网;许可证信息还不明确,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 42,917。

总安装 1,028周安装 42GitHub Stars 32下载 333核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:feature-vision(特征视觉)
来源仓库:https://github.com/steveclarke/dotfiles
仓库路径:skills/feature-vision
安装命令:
npx skills add https://github.com/steveclarke/dotfiles --skill feature-vision
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/steveclarke/dotfiles --skill feature-vision

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

feature-vision 提供结构化命令引导创建愿景文档,系统化梳理功能目标和方向。适用于 Codex、Claude、Cursor、Gemini CLI,适合功能启动前的目标对齐时使用。依据功能开发流程指南构建文档结构,确保各章节内容完整一致。安装前需确认权限范围和维护状态,注意可能触发文件读写和本地文档访问。

适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

来源仓库内路径:skills/feature-vision

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/steveclarke/dotfiles --skill feature-vision;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

feature-vision 怎么安装?

在终端执行 npx skills add https://github.com/steveclarke/dotfiles --skill feature-vision。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

feature-vision 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

feature-vision 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

feature-vision 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

feature-vision 用的人多吗?

公开数据是累计安装 1,028 次,近一周安装 42 次,下载 333 次,GitHub 32 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

feature-vision 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
feature-vision

特征视觉 · 本页

1,02832需要联网未明确通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Vision Document

Overview

Structured command to help create vision documents by guiding through focused discovery phases and building each section systematically following our established feature development process.

Create a vision document following the established feature development process. First, read the feature development guide at @~/.local/share/dotfiles/ai/guides/feature-development-process.md to understand the vision document structure and approach.

Note: If the project has a local copy of the guide at @project/guides/feature-development-process.md, you may reference that instead for project-specific modifications.

IMPORTANT: Start by creating a TODO list to track the 5 phases of vision document creation. Mark the first phase as "in_progress" and others as "pending".

Vision Document Structure

Our vision documents are lean and focused - typically 100-150 lines, never more than 200. They flow from general to specific:

Required sections:

  • Clear vision statement (one sentence)
  • Table of contents (with links to all major sections)
  • Problem (what's broken, who it affects, why it matters)
  • Solution (how we'll solve it, core philosophy, key capabilities)
  • User Needs (simple bullets by role - what they need to accomplish)
  • Future Possibilities (brief overview with reference to separate future.md)
  • Scope (concrete list of what v1 delivers and what's not included)

What does NOT belong in vision:

  • Detailed business justifications or ROI calculations
  • Deployment details (where/how it runs - that's spec territory)
  • Technical implementation details (APIs, schemas, architecture - that's spec territory)
  • Detailed user stories or acceptance criteria (that's requirements territory)
  • Step-by-step workflows or process flows (that's requirements territory)
  • Repetitive sections that restate the same information

Process: Work Through These 5 Phases Sequentially

Phase 1: Problem Discovery

Ask focused questions to understand:

  • What specific problem are we solving?
  • Who experiences this problem and how?
  • What's the current painful/manual process?
  • What are the consequences of not solving this?

ONE QUESTION AT A TIME - Wait for answers before proceeding. NO CORPORATE SPEAK - Keep language conversational and direct.

Phase 2: Solution Exploration

Ask about:

  • How do we envision solving this at a high level?
  • What's our core philosophy or approach?
  • What are the 3-5 key capabilities we need?
  • What makes this solution different/better?

Phase 3: User Needs Analysis

Use consolidated user roles at vision level:

  • End Users (all user types)
  • Admin Staff (consolidate all staff types - Staff, Administrators, Organization Admins)
  • Product Team (if relevant)

Avoid fine-grained role distinctions - detailed role breakdown goes in requirements document.

For each consolidated role, ask ONLY:

  • What do they need to accomplish with this feature?

STOP THERE. Don't ask about "current pain points" or "how success looks different" - that creates redundancy with the Problem section and belongs in requirements.

Phase 4: Future & Scope Discovery

Ask about:

  • What nice-to-have features should we consider for future phases?
  • What scope are we explicitly INCLUDING in v1? (get concrete details)
  • What scope are we explicitly NOT including in v1?

Key: Get detailed, concrete information about what v1 will actually deliver - this becomes the bridge to requirements.

Phase 5: Document Generation

Before generating, ruthlessly eliminate redundancy:

  • Does this section repeat information from another section?
  • Is this section justifying something already obvious from the problem?
  • Is this spec-level detail (deployment, technical architecture)?
  • Is this requirements-level detail (detailed workflows, acceptance criteria)?

Then generate three documents:

1. vision.md - The lean, focused vision document:

  • Create a clear one-sentence vision statement
  • Generate a table of contents with links to all major sections
  • Compile vision document with concrete scope section
  • Flow from general (problems, solution) to specific (detailed scope)
  • Follow "no corporate speak" principle throughout
  • Reference future.md in the brief future possibilities section
  • Keep total length under 150 lines - be ruthless about brevity

2. future.md - Detailed future enhancements:

  • Create separate detailed future.md document with all future features
  • Organize by phases (Phase 2, Phase 3, etc.)
  • Include strategic future vision (if applicable)

3. discussion-summary.md - Living discussion summary document:

  • Create a unified discussion summary that will accumulate across all phases
  • Include "Vision Phase" section with all vision-related discussions
  • Include placeholder sections for Requirements, Spec, and Plan phases
  • Document architectural decisions made and their rationale in Key Decisions Log
  • Capture technical feasibility discussions and implementation ideas
  • Note any technical constraints or challenges identified
  • Include open questions that should be addressed in requirements/spec
  • Add Technical Context section with files referenced and related features
  • Organize with clear sections and table of contents
  • This document serves as institutional memory across the entire feature lifecycle

Key Principles

Flow from General to Specific: Start with high-level problems and solutions, end with concrete scope details that bridge naturally to requirements.

Scope Section is Critical: This should be a concrete LIST of what v1 delivers. Think: bullet points of capabilities/tools/features. This bridges vision to requirements.

BUT don't include:

  • How it will be deployed (staging vs production) - that's spec territory
  • Technical architecture details - that's spec territory
  • Justifications for why things are excluded - just list what's out, no explanations needed

Separate Future Content: Create detailed future.md document for all future features. Vision should only have brief future overview.

No Corporate Speak: Write for humans first. Use clear, simple language. Avoid jargon and unnecessarily complex terminology.

User Role Consolidation: Keep user roles consolidated at vision level. Detailed role breakdowns belong in requirements.

Eliminate Redundancy (CRITICAL): We're a small team that values lean, readable documents. Every section must add NEW information. Before including a section, ask: "Does this repeat something already said?" If yes, DELETE IT.

Common redundant patterns to AVOID:

  • "Business Case" that just restates the problem in corporate language
  • "Consequences" section that repeats pain points already in Problem
  • "Who This Affects" section when it's obvious from Problem
  • Multiple sections listing benefits/value in different ways
  • Detailed "current pain points" AND "how success looks different" in user needs (Problem section already covers this)

Keep It Short: Vision documents should be 100-150 lines MAX. If you're over 200 lines, you're repeating yourself or including spec/requirements-level detail.

Guidelines

  • Ask ONE question at a time and wait for response
  • Update your TODO list as you complete each phase
  • Keep questions conversational, not corporate-speak
  • Focus on immediately useful information, avoid fluff
  • Reference any existing notes or context provided
  • Stay practical - we're a small team that values readable, actionable documents
  • During discovery, it's normal to discuss technical feasibility and architecture - capture all of this
  • Always create three documents at the end: vision.md, future.md, and discussion-summary.md

Getting Started Process

  1. Create your 5-phase TODO list immediately
  2. Ask user: "What feature directory are you working in? Please provide the full path (e.g., project/features/FT033-feature-name)"
  3. Ask user: "If you have existing notes, please tag them with @notes.md or @[filename].md so I can review them"
  4. Validate the directory exists and confirm that's where you'll create vision.md, future.md, and discussion-summary.md
  5. Review any tagged notes files
  6. Begin Phase 1 with the first problem discovery question
  7. Keep questions focused and wait for answers before proceeding to next questions
  8. During conversations, capture technical/architectural discussions even if they don't directly go into vision.md

What to Capture in discussion-summary.md

This document should be structured to accumulate context across all feature development phases. For vision phase, create the full document structure:

Document Structure:

# Feature Name: Discussion Summary

Living document capturing technical discussions, research, decisions, and
context from all phases of feature development.

## Table of Contents
- [Vision Phase](#vision-phase)
- [Requirements Phase](#requirements-phase)
- [Spec Phase](#spec-phase)
- [Plan Phase](#plan-phase)
- [Key Decisions Log](#key-decisions-log)
- [Technical Context](#technical-context)

## Vision Phase
[Vision phase content goes here]

## Requirements Phase
*To be added during requirements phase.*

## Spec Phase
*To be added during spec phase.*

## Plan Phase
*To be added during plan phase.*

## Key Decisions Log
| # | Decision | Rationale | Phase |
|---|----------|-----------|-------|
[Decisions from vision phase]

## Technical Context
### Files Referenced
[Backend and frontend files referenced]

### Related Features
[Related feature references]

Vision Phase Content to Include:

  • Core problem identified and all nuances discussed
  • Solution approach and how it evolved during discussion
  • Technical feasibility discussions
  • Architectural decisions made and their rationale
  • Implementation ideas explored (even if not chosen)
  • Technology choices and why (SDKs, frameworks, protocols)
  • Technical constraints identified
  • Integration points with existing systems
  • Open questions for requirements/spec phases

Key Decisions Log:

  • Number decisions starting from 1
  • Include decision, rationale, and "Vision" as the phase
  • Later phases will continue the numbering

Technical Context:

  • List files referenced during discussions (backend and frontend paths)
  • Note related features that may be affected

Purpose: This document serves as institutional memory across the entire feature lifecycle. Each phase adds its section, and the Key Decisions Log and Technical Context accumulate across all phases.

平台分布

Codex

35.23%
按下载量换算117

Claude

32.6%
按下载量换算109

Cursor

21.06%
按下载量换算70

Gemini CLI

9.05%
按下载量换算30

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills