Agent 生成器
- 作者仓库星标 2,528
- 作者更新于 实时读取
- 作者仓库 agent-academy
- 领域
- AI 智能
- 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
- @microsoft · 未声明 license
- Token 消耗评级
- 中等消耗
- 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
- 无特殊要求
- 文件与系统权限
-
- 只读
- 允许写入 / 修改
- 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: agent-academy-mission-builder
description: > This skill scaffolds new content for the microsoft/agent-academy repository at https://github.…
category: AI 智能
runtime: 无特殊运行时
---
# agent-academy-mission-builder 输出预览
## PART A: 任务判断
- 适用问题:提示词、Agent 工作流、模型评估或自动化推理。
- 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“Phase 0: Gather Intel (Interview the User) / Phase 1: Reconnaissance (Repo Check) / Phase 2: Mission Naming”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提示词、Agent 工作流、模型评估或自动化推理,并标出输入、限制和预期结果。
- **02** 执行计划:优先按“Phase 0: Gather Intel (Interview the User) / Phase 1: Reconnaissance (Repo Check) / Phase 2: Mission Naming”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、会按任务需要访问外部网络、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;会按任务需要访问外部网络;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。
先用一个小任务确认它会围绕“Phase 0: Gather Intel (Interview the User) / Phase 1: Reconnaissance (Repo Check) / Phase 2: Mission Naming”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: agent-academy-mission-builder
description: > This skill scaffolds new content for the microsoft/agent-academy repository at https://github.…
category: AI 智能
source: microsoft/agent-academy
---
# agent-academy-mission-builder
## 什么时候使用
- 把 AI / Agent方向的常用动作沉淀成 Agent 可调用的技能 适合处理AI Agent、提示词、模型评估与自动化推理,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…
- 面向提示词、Agent 工作流、模型评估或自动化推理,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「Phase 0: Gather Intel (Interview the User) / Phase 1: Reconnaissance (Repo Check) / Phase 2: Mission Naming」组织步骤,不把推断写成作者事实。
- 读取文件、写入/修改文件;会按任务需要访问外部网络;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 "agent-academy-mission-builder" {
输入层 -> 用户目标 + 目标文件 + 禁止范围 + 验收标准
上下文层 -> Phase 0: Gather Intel (Interview the User) / Phase 1: Reconnaissance (Repo Check) / Phase 2: Mission Naming
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 | 会按任务需要访问外部网络
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
} Agent Academy Mission Builder
This skill scaffolds new content for the microsoft/agent-academy repository at https://github.com/microsoft/agent-academy. It covers two distinct mission types and handles everything from naming to final markdown output, including badge generation for Special Ops.
Think of this skill like a mission briefing officer: it gets the concept from you, does the reconnaissance (checks the repo, checks for duplicates, maps cross-links), then hands you back a fully briefed document you only need to fill in the "boots on the ground" steps for.
Phase 0: Gather Intel (Interview the User)
Before writing anything, collect the following. Ask in one message, not one question at a time:
Required info:
- Mission type: Course mission (Recruit / Operative / Commander) OR Special Ops?
- Topic / feature: What Copilot Studio capability, integration, or concept does this cover?
- Rough description: What will the learner build or do by the end?
- Difficulty (for Special Ops): Beginner / Intermediate / Advanced (maps to 1 / 2 / 3 stars)
- Prerequisites: What must the learner have done or have access to?
For course missions, also ask:
- Which mission number does this follow? (sets the folder name like
07-) - What is the overall course scenario it fits into? (e.g., Operative = Hiring system)
For Special Ops, also ask:
- Industry or use-case context (healthcare, retail, HR, IT, finance, etc.)? Or "general"?
- Any specific integrations (SharePoint, Teams, Dataverse, Azure, MCP, etc.)?
If the user hasn't given you enough to go on, ask. Otherwise proceed.
Phase 1: Reconnaissance (Repo Check)
Use web_search and web_fetch to:
Check for duplicates: Search
site:github.com/microsoft/agent-academy [topic keyword]andmicrosoft.github.io/agent-academy [topic keyword]. If something very similar exists, flag it and ask the user whether to proceed anyway or pivot the angle.Map existing missions: Fetch https://microsoft.github.io/agent-academy/recruit/ and https://microsoft.github.io/agent-academy/operative/ to get the current mission list. Use this to:
- Assign the correct next mission number for course missions
- Identify missions to cross-link (e.g., if this mission uses MCP, link to Mission 10 MCP)
- Spot curriculum gaps the new mission fills
Check Special Ops listing: Fetch https://microsoft.github.io/agent-academy/special-ops/ to see what already exists and identify tags/categories in use.
Report findings briefly before proceeding. If there's a close duplicate, surface it clearly.
Phase 2: Mission Naming
Agent Academy missions use military/spy operation themed names. Rules:
Course missions: Use
🚨 Mission XX: [Action Verb] + [What They Build]Examples:🚨 Mission 03: Create a Multi-Agent Orchestration SystemThe operation name appears in the briefing paragraph:Operation [Codename]Codenames are dramatic and thematic: "Operation Talent Scout", "Operation Symphony", "Operation MCP Rendezvous". They relate to what the mission accomplishes.Special Ops: The lab title is a punchy phrase, often with a symbol if relevant. Examples:
Microsoft Copilot Studio & MCP,Power Platform CLI MCPThe operation name in the briefing: same dramatic style. The badge bottom banner name is a clever play on the mission topic. See badge section.
Naming formula for operation codenames: [Military/Intelligence Term] + [Thematic Noun]. Examples:
- Adaptive Cards → "Operation Flash Card"
- Event Triggers → "Operation Trip Wire"
- AI Safety → "Operation Safe Harbor"
- Document Generation → "Operation Paper Trail"
- Dataverse Grounding → "Operation Deep Root"
Generate 2-3 name options and let the user pick, or auto-select the best one.
Phase 3: Scaffold the Markdown
Course Mission Template
File: docs/[rank]/[NN]-[slug]/index.md
---
title: "Mission NN: [Title]"
---
# [Emoji] Mission [NN]: [Full Title]
**Welcome, [Rank].** [1-2 sentence scene-setter in mission briefing voice.]
Your assignment, should you choose to accept it, is **[Operation Codename]** —
[one sentence high-level goal].
[2-3 sentences expanding on what they'll build and how it connects to the course scenario.]
## Mission Objective
By the end of this mission, you will have:
- [Concrete deliverable 1]
- [Concrete deliverable 2]
- [Concrete deliverable 3]
## Prerequisites
Before embarking on this mission, ensure you have:
- Completed [Mission N-1: Title](../[folder]/)
- [Any specific access or setup]
> [!TIP]
> [Optional tip for getting access if stuck]
## Background Intel
> [!NOTE]
> [Key concept explanation — 2-4 paragraphs. Explain the why before the how. Include an analogy.]
## Mission Tasks
### Task 1: [Name]
[Brief intro sentence]
1. [Step 1]
2. [Step 2]
3. [Step 3]
> [!NOTE]
> [Callout with context or warning about a common mistake]
### Task 2: [Name]
[... repeat pattern ...]
## Mission Complete
**Mission [NN] is completed!** You now have mastered the following skills:
✅ **[Skill 1]**: [One sentence — what they can now do]
✅ **[Skill 2]**: [One sentence]
✅ **[Skill 3]**: [One sentence]
## Related Missions
- [Mission N-1: Title](../[prev-folder]/) — [Why it connects]
- [Mission N+1: Title](../[next-folder]/) — [What builds on this]
## Further Reading
- [Microsoft docs link]
- [Related Academy content]
Special Ops Template
File: docs/special-ops/[slug]/index.md
---
title: "[Mission Title]"
tags:
- [technology-tag]
- [industry-tag]
difficulty: [1|2|3]
---
# [Emoji] [Mission Title]
> **Difficulty**: [stars] | **Time**: ~[N] min | **Type**: Special Ops

**Welcome, agent.** Your mission — should you choose to accept it — is [one-liner].
## What You'll Build
- [Specific deliverable 1]
- [Specific deliverable 2]
- [Specific deliverable 3]
## Prerequisites
- [Access/license requirement]
- [Tool requirement]
> [!WARNING]
> [Access barrier callout if needed]
## Concept Brief
[2-4 paragraphs on the core concept. Why does it exist? When would you use it?
Include an analogy.]
### Key Terms
| Term | Definition |
|------|------------|
| [Term] | [Plain-language definition] |
## The Scenario
> [!NOTE]
> **Scenario:** [Company] is a [industry] company that needs to [problem]. You are building
> an agent that [solution].
## Mission Tasks
### Task 1: [Name]
1. [Step]
2. [Step]
### Task 2: [Name]
[...]
## Mission Accomplished
Congrats, agent — mission accomplished! Claim your badge:
`https://aka.ms/agent-academy-special-ops/[slug]/form`
## Tags
**Technology**: [tag1], [tag2]
**Industry**: [industry]
**Difficulty**: [stars]
## Related Content
- [Relevant course mission link]
- [Official docs link]
Phase 4: Auto-Draft Background Intel and Scenario
Don't leave these blank. Fill them in using knowledge of Copilot Studio and the feature:
- Write 2-4 paragraphs of background. Voice: energetic, not dry docs.
- Always include an analogy. Existing ones in the repo:
- MCP = "USB-C of AI integration"
- Orchestrator = "symphony conductor"
- Agent = "digital coworker" Find a new one for the new concept.
- For Special Ops: draft the fictional scenario with company name, industry, and problem.
Phase 5: Step-by-Step Drafting
Fill in Mission Tasks based on Copilot Studio knowledge:
- One action per step, clear action verbs (Select, Navigate, Enter, Configure, Test)
- Bold UI element names: Test your agent, Publish, Save
- Code format for: values to type, field names, test messages to send
- Use
[!NOTE],[!TIP],[!WARNING]callouts freely - Where exact steps are unknown, add
<!-- TODO: Add screenshot -->and[FILL IN: describe what needs to happen here]placeholders
Phase 6: Tags and Cross-Links (Special Ops Only)
See references/tags.md for full tag taxonomy. Select 1-3 technology tags and 1-3 industry tags.
Difficulty = number of stars on the badge:
- 1 star: Beginner (single feature, under 30 min, no admin required)
- 2 stars: Intermediate (2-3 concepts, 30-60 min, some config)
- 3 stars: Advanced (multi-step, VSCode, Azure/admin required, 60+ min)
Always cross-link to the most relevant course mission. Check docs/recruit/index.md, docs/operative/index.md, docs/commander/index.md, and docs/special-ops/index.md for the current mission inventory.
Phase 7: Deliver Output
- Scaffolded markdown file create the mission file directly in the repo in the corresponding folder. If it's a recruit course mission, create a folder under
docs/recruit/with the appropriate mission name and place anindex.mdfile inside as well as anassetsfolder for images, for example,docs/recruit/mcp-lab/index.md. Repeat this pattern for operative and commander missions under their respective folders. For Special Ops, create the folder underdocs/special-ops/with the slug name and place theindex.mdfile andassetsfolder inside, for example,docs/special-ops/mcp-lab/index.md. - Summary of what was pre-filled vs. what needs manual work (be explicit about TODOs)
- Cross-link map of which missions link here, and where this links
- For Special Ops: Tags, difficulty, badge banner name, and image generation prompt
- Duplicate check result with links to anything similar that exists
End every delivery with:
"The bones are done. Fill in the step-by-step and swap in real screenshots. Want me to draft any specific task section in more detail?"
Reference Files
Read these when needed:
WRITING_STYLE.md(repo root) — Voice, tone, formatting rulesdocs/recruit/index.md— Recruit mission inventory for numbering and cross-linkingdocs/operative/index.md— Operative mission inventorydocs/commander/index.md— Commander mission inventorydocs/special-ops/index.md— Existing Special Ops for tag patterns and duplicate check
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