Token导航 LogoToken导航

Agent Skill · 研究检索

blueprint-init(蓝图初始化)

blueprint-init 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。

blueprint-init 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,权限为只读;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 41,077。

总安装 1,223周安装 52GitHub Stars 29下载 428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:blueprint-init(蓝图初始化)
来源仓库:https://github.com/laurigates/claude-plugins
仓库路径:skills/blueprint-init
安装命令:
npx skills add https://github.com/laurigates/claude-plugins --skill blueprint-init
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/laurigates/claude-plugins --skill blueprint-init

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。

blueprint-init 用于在当前项目中初始化 Blueprint 开发环境,创建必要的目录结构。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。自动检测 monorepo 上下文,生成 manifest.json 和基础文档框架。

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。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/blueprint-init

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/laurigates/claude-plugins --skill blueprint-in…;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

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

blueprint-init 怎么安装?

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

blueprint-init 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。

blueprint-init 需要什么权限?

权限级别是只读。该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

blueprint-init 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。

blueprint-init 用的人多吗?

公开数据是累计安装 1,223 次,近一周安装 52 次,下载 428 次,GitHub 29 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。

blueprint-init 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。
  • 标注为只读权限,试用阶段不容易误改本地文件。

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。
  • 只想先试试效果、不希望它改动本地文件的人。

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
blueprint-init

蓝图初始化 · 本页

1,22329只读未明确通过

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

Initialize Blueprint Development in this project.

Steps:

  1. Check if already initialized:

- Look for docs/blueprint/manifest.json - If exists, read version and ask user: Use AskUserQuestion: question: "Blueprint already initialized (v{version}). What would you like to do?" options: - "Check for upgrades" → run /blueprint:upgrade - "Reinitialize (will reset manifest)" → continue with step 2 - "Cancel" → exit

1a. Detect monorepo context (format_version 3.3.0+):

  • Walk upward from the current directory looking for an ancestor docs/blueprint/manifest.json (stop at the repo root or $HOME).
  • If an ancestor root manifest exists, this init is creating a child workspace. Capture the relative path from the child back to the root.
  • Additionally scan descendants (max depth 4, skipping node_modules, .git, dist, build, target, .venv) for existing docs/blueprint/manifest.json. If any are found, this init is creating a root that will own existing children.
  • Otherwise this is a standalone blueprint (no workspaces block written).
Use AskUserQuestion (only when ancestor root detected):
question: "Found a parent blueprint at {parent_path}. Register this as a child workspace?"
options:
  - label: "Yes - register as child"
    description: "Writes workspaces.role=child + root_relative_path; root picks it up on next /blueprint:workspace-scan"
  - label: "No - treat as standalone"
    description: "No workspaces block written; this project is independent"
  1. Ask about feature tracking (use AskUserQuestion): question: "Would you like to enable feature tracking?" options: - label: "Yes - Track implementation against requirements" description: "Creates feature-tracker.json to track FR codes from a requirements document" - label: "No - Skip feature tracking" description: "Can be added later with /blueprint:feature-tracker-sync" If "Yes" selected: a. Search for markdown files in the project that contain requirements, features, or user stories b. Auto-detect the most likely source document based on content analysis c. Create docs/blueprint/feature-tracker.json from template using the detected source d. Set has_feature_tracker: true in manifest
  2. Ask about document migration (use AskUserQuestion): Search for existing markdown documentation files across the project (excluding standard files like README.md, CHANGELOG.md, CONTRIBUTING.md, LICENSE.md, CODE_OF_CONDUCT.md, SECURITY.md). # Find markdown files that look like documentation (not standard repo files) find. -name '*.md' -not -path '*/node_modules/*' -not -path '*/.git/*' | grep -viE '(README|CHANGELOG|CONTRIBUTING|LICENSE|CODE_OF_CONDUCT|SECURITY)\.md$' If documentation files found (e.g., REQUIREMENTS.md, ARCHITECTURE.md, DESIGN.md, docs in non-standard locations): Use AskUserQuestion: question: "Found existing documentation: {file_list}. Migrate these to Blueprint-managed paths? (Strongly recommended)" options: - label: "Yes, migrate documents (Recommended)" description: "Move docs into docs/prds/, docs/adrs/, docs/prps/ based on content type. Prevents stale and orphaned documents." - label: "No, leave them in place" description: "Warning: unmigrated docs may become stale or duplicated as Blueprint creates its own documents" If "Yes" selected: a. Analyze each file to determine type: Migrated documentation: - REQUIREMENTS.md → docs/prds/requirements.md - ARCHITECTURE.md → docs/adrs/0001-initial-architecture.md If no documentation files found: Skip this step silently.

- Contains requirements, features, user stories → docs/prds/ - Contains architecture decisions, trade-offs → docs/adrs/ - Contains implementation plans → docs/prps/ - General documentation → docs/ b. Move files to appropriate docs/ subdirectory c. Rename to kebab-case if needed (REQUIREMENTS.md → requirements.md) d. Report migration results:

  1. Ask about maintenance task scheduling (use AskUserQuestion): question: "How should blueprint maintenance tasks run?" options: - label: "Prompt before running (Recommended)" description: "Always ask before running maintenance tasks like sync, validate" - label: "Auto-run safe tasks" description: "Read-only tasks (validate, sync, status) run automatically when due" - label: "Fully automatic" description: "All tasks run automatically on schedule, including writes like rule generation" - label: "Manual only" description: "Tasks only run when you explicitly invoke them" Store selection for task_registry defaults:

- Prompt: all auto_run: false, default schedules - Auto-run safe: read-only tasks (adr-validate, feature-tracker-sync, sync-ids) get auto_run: true; write tasks get false - Fully automatic: all tasks get auto_run: true, default schedules - Manual only: all auto_run: false, all schedules set to on-demand

4a. Ask about generated-rules output path (use AskUserQuestion):

Only prompt when .claude/rules/ already exists and contains files (i.e., hand-authored rules that pre-date blueprint). Skip silently in fresh repos and use the default.

# Only prompt if .claude/rules/ has any content not created by blueprint
find .claude/rules -maxdepth 1 -type f -name '*.md'
Use AskUserQuestion (only when .claude/rules/ has existing content):
question: "Detected existing content in .claude/rules/. Where should blueprint write generated rules?"
options:
  - label: ".claude/rules/blueprint/ (Recommended)"
    description: "Isolated subdirectory — keeps blueprint-managed and hand-authored rules separate, prevents collisions on regenerate"
  - label: ".claude/rules/ (flat)"
    description: "Write generated rules alongside hand-authored ones; risk of overwrite when filenames collide"

Store the chosen path in structure.generated_rules_path in the manifest (defaults to .claude/rules/ when unset). This keeps blueprint-generate-rules and blueprint-derive-rules from clobbering hand-curated rule files (issue #1043).

  1. Ask about decision detection (use AskUserQuestion): question: "Would you like to enable automatic decision detection?" options: - label: "Yes - Detect decisions worth documenting" description: "Claude will notice when conversations contain architecture decisions, feature requirements, or implementation plans that should be captured as ADR/PRD/PRP documents" - label: "No - Manual commands only" description: "Use /blueprint:derive-prd, /blueprint:derive-adr, /blueprint:prp-create explicitly when you want to create documents" Set has_document_detection in manifest based on response. If enabled: Copy document-management-rule.md template to .claude/rules/document-management.md. This rule instructs Claude to watch for:

- Architecture decisions being made during discussion → prompt to create ADR - Feature requirements being discussed or refined → prompt to create/update PRD - Implementation plans being formulated → prompt to create PRP

  1. Create directory structure: Blueprint structure (in docs/blueprint/): docs/ ├── blueprint/ │ ├── manifest.json # Version tracking and configuration │ ├── feature-tracker.json # Progress tracking (if enabled) │ ├── work-orders/ # Task packages for subagents │ │ ├── completed/ │ │ └── archived/ │ ├── ai_docs/ # Curated documentation (on-demand) │ │ ├── libraries/ │ │ └── project/ │ └── README.md # Blueprint documentation ├── prds/ # Product Requirements Documents ├── adrs/ # Architecture Decision Records └── prps/ # Product Requirement Prompts Claude configuration (in.claude/): .claude/ ├── rules/ # Modular rules (including generated) │ ├── development.md # Development workflow rules │ ├── testing.md # Testing requirements │ └── document-management.md # Document organization rules (if detection enabled) └── skills/ # Custom skill overrides (optional)
  2. Create manifest.json (v3.3.0 schema — canonical filename is docs/blueprint/manifest.json, no dot prefix): {"format_version": "3.3.0", "created_at": "[ISO timestamp]", "updated_at": "[ISO timestamp]", "created_by": {"blueprint_plugin": "3.3.0"}, "project": {"name": "[detected from package.json/pyproject.toml or directory name]", "detected_stack": []}, "structure": {"has_prds": true, "has_adrs": true, "has_prps": true, "has_work_orders": true, "has_ai_docs": false, "has_modular_rules": true, "has_feature_tracker": "[based on user choice]", "has_document_detection": "[based on user choice]", "claude_md_mode": "both", "generated_rules_path": "[based on Step 4a; defaults to.claude/rules/ when prompt skipped]"}, "feature_tracker": {"file": "feature-tracker.json", "source_document": "[auto-detected]", "sync_targets": ["TODO.md"]}, "generated": {"rules": {}, "commands": {}}, "custom_overrides": {"skills": [], "commands": []}, "task_registry": {"derive-prd": {"enabled": true, "auto_run": false, "last_completed_at": null, "last_result": null, "schedule": "on-demand", "stats": {}, "context": {}}, "derive-plans": {"enabled": true, "auto_run": false, "last_completed_at": null, "last_result": null, "schedule": "weekly", "stats": {}, "context": {}}, "derive-rules": {"enabled": true, "auto_run": false, "last_completed_at": null, "last_result": null, "schedule": "weekly", "stats": {}, "context": {}}, "generate-rules": {"enabled": true, "auto_run": false, "last_completed_at": null, "last_result": null, "schedule": "on-change", "stats": {}, "context": {}}, "adr-validate": {"enabled": true, "auto_run": "[based on maintenance task choice: true if auto-run safe, false otherwise]", "last_completed_at": null, "last_result": null, "schedule": "weekly", "stats": {}, "context": {}}, "feature-tracker-sync": {"enabled": true, "auto_run": "[based on maintenance task choice: true if auto-run safe, false otherwise]", "last_completed_at": null, "last_result": null, "schedule": "daily", "stats": {}, "context": {}}, "sync-ids": {"enabled": true, "auto_run": "[based on maintenance task choice: true if auto-run safe, false otherwise]", "last_completed_at": null, "last_result": null, "schedule": "on-change", "stats": {}, "context": {}}, "claude-md": {"enabled": true, "auto_run": false, "last_completed_at": null, "last_result": null, "schedule": "on-change", "stats": {}, "context": {}}, "curate-docs": {"enabled": false, "auto_run": false, "last_completed_at": null, "last_result": null, "schedule": "on-demand", "stats": {}, "context": {}}}} Note: Include feature_tracker section only if feature tracking is enabled. Note: As of v3.2.0, progress tracking is consolidated into feature-tracker.json (work-overview.md removed). Monorepo workspaces block (v3.3.0+), appended to the manifest based on the detection from Step 1a:

- Child (ancestor blueprint found and user opted in): "workspaces": {"role": "child", "root_relative_path": "[relative path from this dir to the root]"} - Root (descendant blueprints found): "workspaces": {"role": "root", "discovery_strategy": "auto-cache", "last_scanned_at": null, "children": []} After writing the manifest, run /blueprint:workspace-scan once to populate children[]. - Standalone: omit the workspaces block entirely.

  1. Create initial rules:

- development.md: TDD workflow, commit conventions - testing.md: Test requirements, coverage expectations - document-management.md: Document organization rules (if decision detection enabled)

  1. Handle .gitignore:

- Always commit CLAUDE.md and .claude/rules/ (shared project instructions) - Add docs/blueprint/work-orders/ to .gitignore (task-specific, may contain sensitive details) - If secrets detected in .claude/, warn user and suggest .gitignore entries

  1. Report:
Blueprint Development initialized! (v3.3.0)

Blueprint structure created:
- docs/blueprint/manifest.json
- docs/blueprint/work-orders/
- docs/blueprint/ai_docs/
- docs/blueprint/README.md
[- docs/blueprint/feature-tracker.json (if feature tracking enabled)]

Project documentation:
- docs/prds/           (Product Requirements Documents)
- docs/adrs/           (Architecture Decision Records)
- docs/prps/           (Product Requirement Prompts)

Claude configuration:
- .claude/rules/       (modular rules, including generated)
- .claude/skills/      (custom skill overrides)

Configuration:
- Rules mode: both (CLAUDE.md + .claude/rules/)
[- Feature tracking: enabled]
[- Decision detection: enabled (Claude will prompt when discussions should become ADR/PRD/PRP)]
[- Task scheduling: {prompt|auto-run safe|fully automatic|manual only}]

[Migrated documentation:]
[- {original} → {destination} (for each migrated file)]

Architecture:
- Plugin layer: Generic commands from blueprint-plugin (auto-updated)
- Generated layer: Rules/commands regeneratable from docs/prds/
- Custom layer: Your overrides in .claude/skills/
  1. Prompt for next action (use AskUserQuestion): question: "Blueprint initialized. What would you like to do next?" options: - label: "Derive plans from git history (Recommended)" description: "Analyze commit history, PRs, and issues to build PRDs, ADRs, and PRPs from existing project decisions" - label: "Derive rules from codebase" description: "Analyze commit patterns and code conventions to generate.claude/rules/" - label: "Update CLAUDE.md" description: "Generate or update CLAUDE.md with project context and blueprint integration" - label: "I'm done for now" description: "Exit - you can run /blueprint:status anytime to see options" Based on selection:

- "Derive plans from git history" → Run /blueprint:derive-plans - "Derive rules from codebase" → Run /blueprint:derive-rules - "Update CLAUDE.md" → Run /blueprint:claude-md - "I'm done for now" → Show quick reference and exit

Quick Reference (show if user selects "I'm done for now"):

Management commands:
- /blueprint:status          - Check version and configuration
- /blueprint:upgrade         - Upgrade to latest format version
- /blueprint:derive-prd      - Derive PRD from existing documentation
- /blueprint:derive-adr      - Derive ADRs from codebase analysis
- /blueprint:derive-plans    - Derive docs from git history
- /blueprint:derive-rules    - Derive rules from git commit decisions
- /blueprint:prp-create      - Create a Product Requirement Prompt
- /blueprint:generate-rules  - Generate rules from PRDs
- /blueprint:sync            - Check for stale generated content
- /blueprint:promote         - Move generated content to custom layer
- /blueprint:rules           - Manage modular rules
- /blueprint:claude-md       - Update CLAUDE.md
- /blueprint:feature-tracker-status  - View feature completion stats
- /blueprint:feature-tracker-sync    - Sync tracker with project files

平台分布

Codex

35.81%
按下载量换算153

Claude

28.16%
按下载量换算121

Cursor

17.43%
按下载量换算75

Gemini CLI

8.97%
按下载量换算38

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

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