Agent 生成器
- 作者仓库星标 612
- 作者更新于 实时读取
- 作者仓库 LLMTornado
- 领域
- 文档
- 兼容 Agent
-
- Claude Code
- Cursor
- Cline
- Codex
- Windsurf
- Gemini CLI
- +20
- 信任分
- 85 / 100 · 社区维护
- 作者 / 版本 / 许可
- @lofcz · 未声明 license
- Token 消耗评级
- 低消耗
- 接入复杂程度
- 需简单配置
- 是否需要外部 API Key
- 不需要
- 兼容的系统
- 未声明(默认跨平台)
- 底层运行要求
- Python
- 文件与系统权限
-
- 只读
- 允许写入 / 修改
- Shell 执行
- 网络行为
- 仅限本地
- 安装命令数
- 26 条
档案由构建时根据 SKILL.md 与安装命令自动衍生,可能与作者实际意图存在差异。
需要注意: 未限定 allowed-tools,默认拥有全部工具权限。;上游仓库已 224 天未更新,可能与最新 agent 行为不一致。
---
name: skill-generator
description: Generates complete Anthropic SKILL packages with proper structure, documentation, and automated…
category: 文档
runtime: Python
---
# skill-generator 输出预览
## PART A: 任务判断
- 适用问题:PRD、RFC、README、项目说明或知识库整理。
- 输入要求:目标材料、限制条件、期望输出和验收方式。
- 证据边界:围绕“Trigger Words / Description / Workflow”读取原文规则,不把推断写成作者承诺。
## PART B: 执行结果
- **01** 任务判断:确认你的需求是否属于PRD、RFC、README、项目说明或知识库整理,并标出输入、限制和预期结果。
- **02** 执行计划:优先按“Trigger Words / Description / Workflow”拆成步骤,说明每一步会读取什么、修改什么、产出什么。
- **03** 交付结果:给出可复制的命令、文件改动、检查清单或内容草稿,并说明如何继续迭代。
- **04** 风险边界:结合 读取文件、写入/修改文件、执行终端命令、主要在本地完成、通常不需要额外 API Key 给出执行前确认项。
## Running Rules
- 读取文件、写入/修改文件、执行终端命令;主要在本地完成;通常不需要额外 API Key。
- 先小样例验证,再放大到真实任务。
- 交付时同时给结果、检查口径和下一步迭代建议。 原文出现了 `/tmp` 这类斜杠命令;如果你的 Agent 支持命令触发,优先用命令开场,再补充目标和边界。
告诉 Agent 目标文件或材料、期望结果、不可改范围、是否允许联网或执行命令。本 Skill 的权限画像是:读取文件、写入/修改文件、执行终端命令。
先用一个小任务确认它会围绕“Trigger Words / Description / Workflow”工作;涉及文件或命令时,先看 diff、日志、预览或测试结果。
检查最终产物是否包含明确结果、必要证据和下一步动作;如果输出泛泛而谈,就补充输入、边界和验收标准后重跑。
---
name: skill-generator
description: Generates complete Anthropic SKILL packages with proper structure, documentation, and automated…
category: 文档
source: lofcz/LLMTornado
---
# skill-generator
## 什么时候使用
- 把项目文档方向的常用动作沉淀成 Agent 可调用的技能 适合处理README、PRD、RFC、教程和知识库文档,核心价值是把输入、判断、执行、验证和交付边界固定下来,避免 Agent 泛泛回答。 把任务拆成可执行、可检查、可继续迭代的…
- 面向PRD、RFC、README、项目说明或知识库整理,优先处理能明确输入、步骤和验收标准的工作。
## 需要提供什么
- 目标材料、目录范围、期望结果和不可改动内容。
- 是否允许联网、执行命令、读写文件或调用外部服务。
## 执行规则
- 围绕「Trigger Words / Description / Workflow」组织步骤,不把推断写成作者事实。
- 读取文件、写入/修改文件、执行终端命令;主要在本地完成;通常不需要额外 API Key。
- 先跑小样例,确认结果可检查后再扩大任务范围。
## 输出要求
- 给出最终产物、关键证据、验证方式和下一步动作。
- 信息不足时标记 unknown,不编造命令、平台或依赖。 作者原文负责流程事实;仓库文件负责来源和命令;流狐只补充适用场景、限制和质量判断。
skill "skill-generator" {
输入层 -> 用户目标 + 目标文件 + 禁止范围 + 验收标准
上下文层 -> Trigger Words / Description / Workflow
规则层 -> SKILL.md 触发条件 / 执行顺序 / 输出格式
运行层 -> Python | 读取文件、写入/修改文件、执行终端命令 | 主要在本地完成
安全层 -> 通常不需要额外 API Key + 小任务验证 + diff / 日志复核
输出层 -> 可复制结果 + 检查清单 + 下一步迭代
} Skill Generator - Create Anthropic Skills with Download Verification
Trigger Words
- "create a skill"
- "make a skill"
- "generate a skill"
- "build a skill"
- "new skill"
- "skill generator"
Description
This skill generates complete Anthropic SKILL packages with proper structure, documentation, and automated download verification. It ensures all files are successfully transferred to the user's local machine.
Workflow
Phase 1: Requirements Gathering
- Ask the user for skill specifications:
- Skill name: What should the skill be called?
- Description: What does the skill do?
- Trigger words: What words/phrases activate this skill?
- Functionality: What are the main features?
- Required files: Does it need Python scripts, bash scripts, config files?
Phase 2: Skill Creation
- Create a directory structure:
/tmp/skill-{skillname}/ - Generate the following files:
- SKILL.md: Main skill documentation with:
- Trigger words section
- Description
- Workflow/instructions
- Usage examples
- Technical details
- README.md: User-facing documentation
- Any additional files needed (Python scripts, bash scripts, configs, etc.)
- SKILL.md: Main skill documentation with:
Phase 3: File Verification (Pre-Download)
- List all created files in
/tmp/skill-{skillname}/ - Display file count and names to user
- Ask user to confirm files look correct before download
Phase 4: Download to User Machine
- Copy all files from
/tmp/skill-{skillname}/to$OUTPUT_DIR/skill-{skillname}/ - Maintain directory structure during copy
- Display success message with file list
Phase 5: Post-Download Verification
- Ask user to verify files are on their local machine:
- "Please check your downloads folder for the 'skill-{skillname}' directory"
- "Can you confirm you see the following files: [list files]"
- Wait for user confirmation
- If user confirms success:
- Provide installation instructions
- Explain how to use the skill
- If user reports missing files:
- Troubleshoot the issue
- Offer to regenerate and re-download specific files
Phase 6: Installation Guidance
- Provide clear instructions:
To install this skill: 1. Move the 'skill-{skillname}' folder to your skills directory 2. Ensure the folder structure is: skills/skill-{skillname}/SKILL.md 3. Test the skill by using one of these trigger phrases: {list triggers}
File Structure Template
Every skill should follow this structure:
skill-{skillname}/
├── SKILL.md # Main skill documentation (REQUIRED)
├── README.md # User-facing documentation (RECOMMENDED)
├── examples/ # Example files (OPTIONAL)
├── scripts/ # Python/Bash scripts (OPTIONAL)
└── configs/ # Configuration files (OPTIONAL)
SKILL.md Template
# {Skill Name} - {Brief Description}
## Trigger Words
- "trigger phrase 1"
- "trigger phrase 2"
- "trigger phrase 3"
## Description
{Detailed description of what the skill does}
## Workflow
1. Step 1
2. Step 2
3. Step 3
## Usage Examples
**Example 1:**
User: "{example user message}"
Assistant: {what the assistant should do}
## Technical Details
- {Any technical requirements}
- {Dependencies or prerequisites}
- {Special considerations}
## Files Included
- **SKILL.md**: This file
- **{other files}**: {description}
## Notes
{Any additional notes or warnings}
Implementation Notes
Creating Directory Structure
mkdir -p /tmp/skill-{skillname}/scripts
mkdir -p /tmp/skill-{skillname}/examples
Copying to Output
# Copy entire directory structure to OUTPUT_DIR
cp -r /tmp/skill-{skillname} $OUTPUT_DIR/
File Verification
# Before download - show what will be downloaded
echo "Files ready for download:"
find /tmp/skill-{skillname} -type f -exec basename {} \;
# Count files
find /tmp/skill-{skillname} -type f | wc -l
Best Practices
- Always ask for requirements first - Don't assume what the user needs
- Keep SKILL.md clear and concise - This is what the AI reads
- Make README.md user-friendly - This is for humans
- Include examples - Show how the skill should be used
- Verify before download - Prevent unnecessary re-downloads
- Confirm after download - Ensure files reached the user
- Provide clear installation steps - Don't leave users guessing
Error Handling
If files don't download:
- Check OUTPUT_DIR is accessible
- Verify files exist in /tmp
- Try copying individual files
- Offer to create a zip archive as backup
If user can't find files:
- Ask where they're checking
- Explain typical download locations
- Provide file names to search for
- Offer to regenerate if needed
Example Interaction
User: "Create a skill for processing CSV files"
Assistant: "I'll help you create a CSV processing skill! Let me gather some details:
- What should this skill be called? (e.g., 'csv-processor')
- What specific CSV operations should it perform? (e.g., filtering, merging, transforming)
- Should it include any Python scripts for data manipulation?
- Any specific trigger phrases you'd like? (I'll suggest: 'process csv', 'csv tools', 'analyze csv')
Please provide these details so I can create the perfect skill for you!"
Security Considerations
- Validate all user inputs for file names (no special characters that could break paths)
- Keep all operations within allowed directories
- Don't include sensitive information in skill files
- Sanitize any code examples to prevent injection
Maintenance
When updating this skill:
- Test the complete workflow end-to-end
- Verify download mechanism works
- Ensure verification steps are clear
- Update examples if workflow changes
先判断是否适合
作者设计意图
作者的方法与取舍
边界和复核