Agent 生成器
- 作者仓库星标 6,311
- 作者更新于 实时读取
- 作者仓库 OpenSpace
- 领域
- 数据
- 兼容 Agent
-
- Claude Code
- Cursor
- Cline
- Codex
- Windsurf
- Gemini CLI
- +20
- 信任分
- 88 / 100 · 社区维护
- 作者 / 版本 / 许可
- @HKUDS · 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
- 需要 · Vendor-specific
- 兼容的系统
- 未声明(默认跨平台)
- 底层运行要求
- 无特殊要求
- 文件与系统权限
-
- 只读
- 允许写入 / 修改
- Shell 执行
- 网络行为
- 允许外网请求
- 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。
---
name: skill-template-generator
description: Generate properly-formatted SKILL.md files from extracted architectural patterns. Turns raw patt…
category: 数据
runtime: 无特殊运行时
---
# skill-template-generator 输出预览
## PART A: 任务判断
- 适用问题:表格、CSV、数据集、指标或分析流程。
- 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“When to Use / SKILL.md Format / Frontmatter (Required)”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表格、CSV、数据集、指标或分析流程,并标出输入、限制和预期结果。
- **02** 执行计划:优先按“When to Use / SKILL.md Format / Frontmatter (Required)”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、会按任务需要访问外部网络、需要准备 Vendor-specific 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;会按任务需要访问外部网络;需要准备 Vendor-specific 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。 原文没有稳定的斜杠命令要求。安装验证后通常全局生效,直接在对话里点名这个 Skill 并描述任务即可。
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“When to Use / SKILL.md Format / Frontmatter (Required)”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-template-generator
description: Generate properly-formatted SKILL.md files from extracted architectural patterns. Turns raw patt…
category: 数据
source: HKUDS/OpenSpace
---
# skill-template-generator
## 什么时候使用
- 把数据处理方向的常用动作沉淀成 Agent 可调用的技能 适合处理表格、CSV、指标、数据集、分析和可视化报告,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、可继续迭代的步…
- 面向表格、CSV、数据集、指标或分析流程,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「When to Use / SKILL.md Format / Frontmatter (Required)」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;会按任务需要访问外部网络;需要准备 Vendor-specific 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-template-generator" {
输入层 -> 用户目标 + 目标文件 + 禁止范围 + 验收标准
上下文层 -> When to Use / SKILL.md Format / Frontmatter (Required)
规则层 -> SKILL.md 触发条件 / 执行顺序 / 输出格式
运行层 -> 无特殊运行时 | 读取文件、写入/修改文件、执行终端命令 | 会按任务需要访问外部网络
安全层 -> 需要准备 Vendor-specific 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 Template Generator
Given a structured pattern description (from codebase-pattern-analyzer or manual analysis), generate a valid SKILL.md file that OpenSpace can register, select for tasks, and evolve over time.
When to Use
- After extracting a pattern from a reference codebase, you need to package it as a reusable skill
- You want to create a new skill that teaches an agent how to perform a specific kind of task
- You are converting informal documentation or code examples into the SKILL.md format
SKILL.md Format
Every skill is a directory containing at minimum a SKILL.md file:
my-skill-name/
├── SKILL.md # Required — the skill definition
├── scripts/ # Optional — helper scripts
├── references/ # Optional — reference data files
└── assets/ # Optional — images, templates
Frontmatter (Required)
The file MUST start with YAML frontmatter containing exactly two fields:
---
name: my-skill-name
description: One-sentence description of what this skill teaches. Must be specific enough for LLM skill selection to match it to relevant tasks.
---
Rules:
namemust match the directory name (kebab-case, lowercase)descriptionshould be 15-40 words, include key terms an LLM would search for- No other frontmatter fields — everything else goes in the markdown body
Body Structure
The markdown body follows this template:
# [Skill Title]
[1-2 sentence summary of what this skill enables]
## When to Use
- [Trigger condition 1]
- [Trigger condition 2]
- [Trigger condition 3]
## [Core Content Sections]
[Step-by-step instructions, code templates, API references, etc.]
## Key Patterns
1. [Convention or best practice 1]
2. [Convention or best practice 2]
...
Step 1: Determine Skill Category
Match the pattern to one of three OpenSpace categories:
| Category | Use When | Example |
|---|---|---|
tool_guide |
The skill teaches how to use a specific tool or technique | "How to analyze a codebase", "How to use Finnhub API" |
workflow |
The skill prescribes an end-to-end procedure with ordered steps | "Create a panel component", "Set up API proxy endpoint" |
reference |
The skill provides knowledge that informs decisions | "WorldMonitor architecture index", "News API comparison" |
The category affects how OpenSpace judges the skill during execution analysis:
- workflow: Was the agent able to follow the prescribed steps?
- tool_guide: Did the agent use the described tool/approach?
- reference: Did the knowledge influence agent decisions?
Step 2: Write the Description
The description field is the most critical line — it determines whether OpenSpace selects this skill for a given task.
Good descriptions (specific, searchable):
- "Create a dashboard panel component using vanilla TypeScript DOM API, following the worldmonitor Panel architecture."
- "Integrate with Finnhub Stock API for real-time and historical stock market data."
- "CSS grid layout system for a responsive panel dashboard with dark theme."
Bad descriptions (vague, generic):
- "A useful skill for building things."
- "TypeScript patterns."
- "How to make a panel."
Formula: [Action verb] + [specific subject] + [key technology/approach] + [context/project reference]
Step 3: Write Actionable Instructions
The body must be concrete enough that an AI agent can follow it without guessing.
For Workflow Skills
Include:
- File paths — exact paths where files should be created (
src/components/MyPanel.ts) - Code templates — complete, runnable code blocks (not pseudocode)
- Import statements — every import the code needs
- Interface definitions — typed data shapes
- Integration points — how this connects to other parts of the project
Use {baseDir} for paths relative to the skill directory:
Read the reference file at {baseDir}/references/example.json
For Tool Guide Skills
Include:
- API endpoints — full URLs with parameter documentation
- Auth mechanism — how to authenticate (query param, header, etc.)
- Response shapes — JSON examples with field descriptions
- Rate limits — free tier limitations
- Error handling — common errors and how to handle them
For Reference Skills
Include:
- Structured index — tables, lists, or maps of the reference material
- Key file paths — where to find specific things in the reference codebase
- Architecture decisions — WHY certain patterns were chosen
- Comparison tables — alternatives and tradeoffs
Step 4: Add Dependency Hints
If the skill depends on patterns from other skills, mention them explicitly:
## Prerequisites
This skill builds on:
- `panel-component` — for the Panel base class
- `data-service` — for the circuit breaker pattern
- `api-proxy-endpoint` — for server-side API key isolation
This helps OpenSpace understand skill composition when DERIVING new skills.
Step 5: Validate the Skill
Before saving, verify:
- Frontmatter is valid YAML — no tabs, proper indentation, quotes around special chars
- Name matches directory —
name: foo-barlives infoo-bar/SKILL.md - Code blocks are complete — every snippet can be copy-pasted and run
- No broken references — all file paths and URLs are valid
- Description is specific — would an LLM pick this skill for the right task?
Example: Converting a Pattern to a Skill
Input (from codebase-pattern-analyzer):
Pattern: Circuit Breaker Data Service
Source: worldmonitor/src/services/*.ts
Category: service
Structure: Module-level CircuitBreaker instance, async fetch functions, typed interfaces
Key code: createCircuitBreaker({ name, cacheTtlMs }), breaker.execute(fn, default)
Output (data-service/SKILL.md):
---
name: data-service
description: Create data fetching services with circuit breaker pattern for API resilience. Services handle fetch, cache, retry, and expose typed data to panel components.
---
# Data Service Pattern
Each panel's data comes from a dedicated service module in `src/services/`. ...
## Circuit Breaker
[Full implementation code]
## Service Module Pattern
[Template with typed interfaces, breaker usage, export functions]
## Key Patterns
1. One circuit breaker per API endpoint
2. Export typed interfaces for data shapes
3. Wrap fetch calls in breaker.execute(fn, defaultValue)
...
Evolution Hooks
Skills generated by this workflow are designed to evolve:
- FIX: When an API changes or code pattern breaks, OpenSpace updates the skill in-place
- DERIVED: When a new panel needs a similar service, OpenSpace derives a specialized version
- CAPTURED: When an agent discovers a novel pattern during execution, OpenSpace captures it as a new skill
The better the initial skill quality, the better the evolved descendants.
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