SuperAPI AI Directory

tech-survey

Writing & research

View repository

Updated 10-15 23:42

About this skill

Tech Survey — Technology Panoramic Survey

Enter a technology name to automatically complete a panoramic survey and output a structured Markdown document with 11 dimensions.

When to Use

  • "调研 WebAssembly" / "调研 Kafka"
  • "技术调研 Kubernetes" / "了解 Rust 技术"
  • "survey gRPC" / "React 技术全景"

Note: Not applicable to everyday Q&A scenarios such as "What is XX", nor to technology selection with a specific project (please use /tech-selection-research)。


Phase 0: Parse Input

Extract the following variables from the user message; they are needed in all subsequent phases:

  • {tech} — Technology name (e.g., "WebAssembly", "Kafka")
  • {tech_en} — English lowercase identifier, derivation rules see below
  • {date} — Current date, format YYYY-MM-DD

{tech_en} Derivation rules (by priority):

  1. Prefer the technology's official lowercase brand name (e.g., webassembly、kubernetes、rust)
  2. If there is no clear official brand name, use lowercase hyphen-separated (e.g., web-assembly)
  3. If the technology has a widely used abbreviation, the abbreviation may be used (e.g., wasm、k8s) — only when the abbreviation is the technology's primary identifier
  4. For Chinese technology names (e.g., "鸿蒙OS"), first use one WebSearch to look up the official English name; when there are multiple spellings, choose the most widespread (e.g., harmonyos)

If the English name cannot be determined, use one WebSearch to quickly look it up; do not skip.

Identification of overly broad technologies: If the input is a broad technology category rather than a specific technology (e.g., "AI", "cloud computing", "distributed systems"), stop execution and prompt:

您输入的「{tech}」范围较广,请确认是否调研特定技术,例如:
- {示例1}、{示例2}、{示例3}
请重新输入更具体的技术名称。

Output file path:{tech_en}-survey-{date}.md, saved in the current working directory.


Phase 1: Pre-survey Agent (main thread, serial)

Dispatch 1 sub-agent to produce a "technology snapshot" shared by all agents in Phase 2, ensuring consistency of key data such as version numbers and competitors.

description: "预调研{tech}技术快照"

prompt:

你是技术调研专家。请对「{tech}」进行快速预调研,产出技术快照。

执行 4–6 次 WebSearch,覆盖以下内容:
1. 当前稳定版本 / 最新版本
2. 主要竞品 / 相关技术(3–5 个)
3. 诞生年份、主导组织/公司
4. 核心卖点一句话
5. 治理模式(开放标准 / 开源项目 / 商业产品)

使用 WebFetch 访问至少 1 个官方页面(官网、GitHub 首页、版本发布页)验证版本信息。

输出格式(纯文本,600 字符上限;来源链接如空间紧张可合并为一行):

【技术快照:{tech}】
当前版本:xxx
诞生时间:xxxx年
主导方:xxx
治理模式:开放标准(W3C)/ 开源项目(Apache 2.0)/ 商业产品(公司名)
核心定位:一句话
主要竞品/相关技术:A、B、C
来源链接:[url1] [url2]

If snapshot collection fails, print a warning in the terminal and continue to Phase 2 (each Agent searches for version information on its own, without injecting the snapshot).


Phase 2: 5 Parallel Sub-agents

Key requirement: All 5 Agents must be placed inthe same messageto ensure parallel execution.

Each Agent is injected with the technology snapshot produced in Phase 1 as context.

Each Agent uses subagent_type: "general-purpose"。

language strategy: For technologies whose main community is English (e.g., WebAssembly, Kubernetes, Rust), supplement Chinese search keywords with English keywords. For Chinese local technologies or technologies mainly documented in Chinese, prioritize Chinese searches.

WebFetch requirement: Each Agent must use WebFetch to visit at least 1 search result page to verify data accuracy. Source links must be actually visited URLs; placeholders must not be used.


Agent 1 — Chapters 1 & 2 (Origin · Positioning · Use Cases + Evolution Timeline)

description: "调研{tech}起源定位与演进时间线"

prompt:

你是技术调研专家。请为「{tech}」采集以下内容,用于生成调研文档的第一、二章节。

【技术快照(请以此为基准,避免与之矛盾)】
{tech_snapshot}

执行 3–5 次 WebSearch,至少使用 WebFetch 访问 1 个页面验证数据:

搜索方向:
- "{tech}" 诞生原因 历史 背景
- "{tech}" history origin background(英语主社区技术补充)
- "{tech}" vs {竞品} 定位对比
- "{tech}" 使用场景 适合 不适合 use cases
- "{tech}" 版本历史 timeline changelog

输出(中间笔记,1000 字符上限,全部中文):

**【章节一 起源·定位·使用场景】**
产生背景:{解决什么问题,前驱技术是什么}
定位差异:{与快照中竞品的区别,可用简表}
使用场景:🟢已成熟:{场景} | 🟡增长:{场景} | 🔴不适合:{场景}

**【章节二 演进时间线】**
{年份} → {里程碑}
{年份} → {里程碑}
...

**来源链接**(实际访问过的 URL):
- {url}

Agent 2 — Chapter 3 (Core Concept System)

description: "调研{tech}核心概念体系"

prompt:

你是技术调研专家。请为「{tech}」采集核心概念体系内容,用于生成调研文档的第三章节。

【技术快照(请以此为基准,避免与之矛盾)】
{tech_snapshot}

执行 3–4 次 WebSearch,至少使用 WebFetch 访问 1 个官方文档/规范页面:

搜索方向:
- "{tech}" 核心概念 架构 architecture
- "{tech}" 关键规范 标准 specification
- "{tech}" 工作原理 how it works

输出(中间笔记,800 字符上限,全部中文):

**【章节三 核心概念体系】**
架构层次:{ASCII 图或简表描述层次关系}
关键规范/组件:
| 名称 | 状态 | 说明 |
|------|------|------|
| ... | ... | ... |
核心概念解释:
1. {概念1}:{定义}
2. {概念2}:{定义}

**来源链接**(实际访问过的 URL):
- {url}

Agent 3 — Chapters 4 & 5 (Runtime Ecosystem + Language/Platform Support)

description: "Research the {tech} runtime ecosystem and language support"

prompt:

你是技术调研专家。请为「{tech}」采集运行时生态和语言/平台支持内容,用于生成调研文档的第四、五章节。

【技术快照(请以此为基准,避免与之矛盾)】
{tech_snapshot}

执行 4–5 次 WebSearch,至少使用 WebFetch 访问 1 个页面验证框架/语言对比数据:

搜索方向:
- "{tech}" 主流框架 运行时 对比 runtime comparison
- "{tech}" 语言支持 language support
- "{tech}" 平台支持 platform

输出(中间笔记,1200 字符上限,全部中文):

**【章节四 运行时生态】**
主流运行时/框架:
| 名称 | 主导方 | 特点 | 适用场景 |
|------|--------|------|---------|
| ... | ... | ... | ... |

**【章节五 语言/平台支持】**
| 语言 | 成熟度 | 工具链 | 说明 |
|------|--------|--------|------|
| ... | ⭐⭐⭐⭐⭐ | ... | ... |

**来源链接**(实际访问过的 URL):
- {url}

Agent 4 — Chapters 6 and 7 (Application scenario panorama + Relationship with related technologies)

description: "Research {tech} application scenarios and comparison with related technologies"

prompt:

你是技术调研专家。请为「{tech}」采集应用场景和与相关技术的关系,用于生成调研文档的第六、七章节。

【技术快照(请以此为基准,快照中的竞品/相关技术作为对比对象)】
{tech_snapshot}

执行 3–5 次 WebSearch,至少使用 WebFetch 访问 1 个页面验证案例数据:

搜索方向:
- "{tech}" 应用案例 生产环境 production use case
- "{tech}" vs {竞品1}(从快照取)
- "{tech}" vs {竞品2}(从快照取)
- "{tech}" 与 {相关技术} 关系 区别

输出(中间笔记,800 字符上限,全部中文):

**【章节六 应用场景全景】**
| 场景 | 典型案例 | 效果/原因 |
|------|---------|----------|
| ... | ... | ... |

**【章节七 与{相关技术}的关系】**(用最相关的 1–2 个技术作为章节标题的填充词)
| 维度 | {tech} | {相关技术} |
|------|--------|-----------|
| ... | ... | ... |
一句话总结:{互补还是竞争,各自适用场景}

**来源链接**(实际访问过的 URL):
- {url}

Agent 5 — Chapters 8, 9, 10, and 11 (Tools + Adoption data + Learning path + Trends)

description: "Research {tech} toolchain, adoption data, learning path, and trends"

prompt:

你是技术调研专家。请为「{tech}」采集工具链、采用数据、学习路径和趋势预判,用于生成调研文档的第八至十一章节。

【技术快照(请以此为基准,避免与之矛盾)】
{tech_snapshot}

执行 5–6 次 WebSearch,至少使用 WebFetch 访问 1 个页面验证采用数据或趋势信息:

搜索方向:
- "{tech}" 核心工具 工具链 toolchain
- "{tech}" 采用数据 使用率 statistics survey
- "{tech}" 入门教程 学习路线 getting started
- "{tech}" 权威资源 书籍 官方文档
- "{tech}" 2025 2026 趋势 roadmap future

输出(中间笔记,1500 字符上限,全部中文):
如字符不足,优先保留工具列表和趋势预判,学习路径可压缩为条目,权威资源表可缩减至 3–4 项。

**【章节八 核心工具与框架】**
| 工具 | 用途 |
|------|------|
| ... | ... |

**【章节九 生态规模与采用数据】**
{渗透率、知名用户、标准组织状态等,附数据来源}

**【章节十 学习路径】**
入门({N}周):{条目列表}
进阶({N}周):{条目列表}
权威资源:
| 资源 | 说明 |
|------|------|
| ... | ... |

**【章节十一 趋势预判】**
| 趋势 | 可信度 | 说明 |
|------|--------|------|
| ... | 高/中/低 | ... |

**来源链接**(实际访问过的 URL):
- {url}

Phase 3: Integration and output (main thread)

After all 5 sub-agents return, the main thread executes the following steps.

Step 1: Consistency check and deduplication

  • Use the version numbers/years in the snapshot to verify each Agent's output; when contradictions are found, the snapshot takes precedence
  • When the same tool/framework appears in multiple chapters, keep it in the most relevant chapter and mark the other chapters with "see Section X"
  • Check the completeness of all source URLs; mark missing ones as "source to be supplemented"
  • If a sub-agent fails, mark the corresponding chapter as "insufficient data, manual supplementation recommended"

Step 2: Write the core conclusions

Based on the 5 Agent outputs, distill 3–5 core conclusions in a fixed format:

| # | 结论 | 行动建议 |
|---|------|---------|
| 1 | {核心发现} | {具体建议} |

Step 3: Use the Write tool to write the file

File path:{tech_en}-survey-{date}.md(current working directory)

Expansion requirements: Write the complete Markdown content based on each Agent's condensed notes; the intermediate notes must not be pasted verbatim as chapter content. Each chapter must be expanded into paragraphs, lists, or tables, with substantial content. Chapter length is not constrained by the Agent character limit.

Output template:

# {tech} 技术全景调研

> 调研日期:{date} | 技术快照:{当前版本} 时代

## 核心结论

| # | 结论 | 行动建议 |
|---|------|---------|
| 1 | ... | ... |

---

## 一、产生原因·定位·使用场景

### 1.1 产生原因

{内容}

### 1.2 技术定位

{内容}

### 1.3 使用场景速查

{内容}

---

## 二、技术演进时间线

{内容}

---

## 三、核心概念体系

{内容}

---

## 四、运行时生态

{内容}

---

## 五、语言/平台支持

{内容}

---

## 六、应用场景全景

{内容}

---

## 七、与 {相关技术} 的关系

{内容}

---

## 八、核心工具与框架

{内容}

---

## 九、生态规模与采用数据

{内容}

---

## 十、学习路径

{内容}

---

## 十一、趋势预判

{内容}

---

*Sources:*
{按章节分组的所有来源链接}

Chapter 7 is the only dynamic title, filled in according to the related technologies identified in Agent 4's output (e.g., "Relationship with containers", "Relationship with REST API"). The remaining chapter titles remain fixed. The final document is expected to be 400–800 lines.

Step 4: Terminal output

✅ {tech} 技术全景调研完成
📄 文件:{tech_en}-survey-{date}.md
📝 章节:11 个 | 来源链接:{N} 个

Error Handling

Situation Handling method
Input too broad Phase 0 stops and prompts the user to narrow the scope, giving 3 specific technology examples
Technology name unrecognizable Prompt the user to confirm, giving a list of common technologies (Kubernetes, React, Kafka, Rust, gRPC, etc.)
Phase 1 snapshot fails Degrade to proceeding directly to Phase 2 without a snapshot; each Agent searches for version information on its own
A Phase 2 Agent fails Mark the corresponding chapter as "insufficient data, manual supplementation recommended", without affecting other chapters
Technology ecosystem is small (emerging technology) Each Agent automatically reduces the relevant tables, and Phase 3 notes in the core conclusions block at the beginning of the document that "this technology ecosystem is still developing, and the data may be incomplete"