survey

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

survey - Cross-Platform Problem-Space Scan

survey - 跨平台问题领域调研

Keyword:
survey
| Platforms: OMC / OMX / OHMG / Claude Code / Codex CLI / Gemini CLI / OpenCode
Survey the landscape before planning, coding, or committing to a direction.

关键词：
survey
| 支持平台：OMC / OMX / OHMG / Claude Code / Codex CLI / Gemini CLI / OpenCode
在规划、编码或确定方向前开展格局调研。

When to use this skill

何时使用此Skill

Before building a new feature, tool, workflow, or agent capability
When the user asks "what exists?", "scan the landscape", "research this space", or "survey solutions"
When you need problem context, current workarounds, and solution gaps before
```
/plan
```
,
```
jeo
```
,
```
ralph
```
, or implementation
When the topic spans multiple agent platforms and you need a single vendor-neutral picture

构建新功能、工具、工作流或Agent能力之前
用户询问“已有哪些方案？”、“调研格局”、“研究该领域”或“调研解决方案”时
在使用
```
/plan
```
、
```
jeo
```
、
```
ralph
```
或开始实施前，需要了解问题背景、当前临时解决方案和方案差距时
主题涉及多个Agent平台，需要一份中立的全景视图时

Do not use this skill when

何时不使用此Skill

The user already knows the solution and wants implementation now
The task is a small bug fix or narrow code change
The user needs a feasibility study, architecture plan, or execution roadmap rather than discovery

用户已经明确解决方案，希望直接开始实施时
任务是小bug修复或范围狭窄的代码变更时
用户需要可行性研究、架构规划或执行路线图，而非探索性调研时

Output Package

输出成果包

Save research in a platform-neutral directory so Claude / Codex / Gemini can all reuse it:

text

.survey/{slug}/
├── triage.md
├── context.md
├── solutions.md
└── platform-map.md    # optional; required for agent/tooling/platform topics

```
triage.md
```
: problem / audience / why now
```
context.md
```
: workflows, affected users, workarounds, adjacent problems, user voices
```
solutions.md
```
: solution inventory, categories, frequency, gaps, contradictions, key insight
```
platform-map.md
```
: normalize platform-specific findings into
```
settings
```
,
```
rules
```
,
```
hooks
```

将研究内容保存到平台无关的目录中，以便Claude / Codex / Gemini均可复用：

text

.survey/{slug}/
├── triage.md
├── context.md
├── solutions.md
└── platform-map.md    # 可选；Agent/工具/平台类主题必填

```
triage.md
```
：问题/受众/当前必要性
```
context.md
```
：工作流、受影响用户、临时解决方案、关联问题、用户反馈
```
solutions.md
```
：方案清单、分类、提及频率、差距、矛盾点、核心洞察
```
platform-map.md
```
：将平台特定发现标准化为
```
settings
```
、
```
rules
```
、
```
hooks
```
三层

Core Abstraction Layer

核心抽象层

When the topic involves agent tooling, model orchestration, hooks, permissions, or vendor APIs, normalize findings into three layers:

```
settings
```
: model, safety, temperature, max tokens, system prompt, provider parameters
```
rules
```
: allow/deny/modify policy, guardrails, action constraints, approval logic
```
hooks
```
: pre/post/error callbacks, notify handlers, lifecycle automation, event triggers

Use this abstraction when comparing Claude, Codex, Gemini, OMC, OMX, or OHMG. Do not describe vendor features as unrelated one-off concepts if they map cleanly into one of these three layers.

Recommended internal profile:

yaml

survey_profile:
  settings:
    search_language: English
    output_language: user-language
    max_searches_per_lane: 8-10
    save_root: .survey
  rules:
    - facts_only
    - no_code_generation
    - ask_max_2_triage_questions
    - dedupe_solution_names
    - include_source_links
  hooks:
    pre_research:
      - normalize_topic
      - choose_platform_adapter
    post_lane:
      - merge_notes
      - record_sources
    post_run:
      - write_summary
      - suggest_next_skill

当主题涉及Agent工具、模型编排、钩子、权限或厂商API时，将发现标准化为以下三层：

```
settings
```
：模型、安全配置、温度系数、最大令牌数、系统提示词、供应商参数
```
rules
```
：允许/拒绝/修改策略、防护规则、操作约束、审批逻辑
```
hooks
```
：前置/后置/错误回调、通知处理器、生命周期自动化、事件触发器

对比Claude、Codex、Gemini、OMC、OMX或OHMG时使用此抽象层。如果厂商功能可以清晰映射到这三层中的某一层，不要将其描述为无关的独立概念。

推荐的内部配置文件：

yaml

survey_profile:
  settings:
    search_language: English
    output_language: user-language
    max_searches_per_lane: 8-10
    save_root: .survey
  rules:
    - facts_only
    - no_code_generation
    - ask_max_2_triage_questions
    - dedupe_solution_names
    - include_source_links
  hooks:
    pre_research:
      - normalize_topic
      - choose_platform_adapter
    post_lane:
      - merge_notes
      - record_sources
    post_run:
      - write_summary
      - suggest_next_skill

Platform Adapter

平台适配层

Select the closest available orchestration surface, but keep the output format identical.

Platform	Preferred execution shape	Notes
OMC / Claude Code	4 parallel research agents. Prefer research-oriented specialist agents when available; otherwise use general-purpose agents with web search.	Keep artifacts in `.survey/` , not `.omc/` , so other platforms can reuse them.
OMX / Codex CLI	Use `$research` , explorer/planner-style agents, or parallel workers with web access.	Translate vendor config into `settings/rules/hooks` in `platform-map.md` when relevant.
OHMG / Gemini / Antigravity	Use Workflow Guide + PM/research-capable agents or equivalent parallel lanes.	Preserve the same 4-lane structure and shared artifact names.
OpenCode / fallback	Use generic web-capable agents in parallel.	Do not block on missing specialized agent names.

If a platform-specific specialist does not exist, fall back to a general-purpose web-enabled agent and keep going.

选择最匹配的编排界面，但保持输出格式一致。

平台	推荐执行方式	说明
OMC / Claude Code	4个并行调研Agent。优先使用面向调研的专业Agent；若不可用则使用带网页搜索的通用Agent。	将成果保存至 `.survey/` 而非 `.omc/` ，以便其他平台复用。
OMX / Codex CLI	使用 `$research` 、探索者/规划者类型Agent，或带网页访问权限的并行工作进程。	相关时将厂商配置转换为 `settings/rules/hooks` 并写入 `platform-map.md` 。
OHMG / Gemini / Antigravity	使用工作流向导 + 具备PM/调研能力的Agent或等效并行通道。	保留相同的4通道结构和共享成果文件名。
OpenCode / fallback	使用带网页访问权限的通用并行Agent。	不要因缺少专业Agent名称而停滞。

如果不存在平台特定的专业Agent，回退到带网页访问权限的通用Agent并继续执行。

Execution Policy

执行规则

Never write product code in this skill. This is a research-only skill.
Search broadly in English for coverage, unless the user explicitly requires another search language.
Save files in the user's language unless the repository has a stronger convention.
Ask at most 2 triage questions, one at a time, only if
```
what
```
,
```
who
```
, or
```
why
```
is unclear.
Run all 4 research lanes in parallel whenever possible.
Keep claims source-backed. Include links for quotes, rankings, and non-obvious claims.
Deduplicate tools that appear under multiple names or product tiers.
Do not recommend build/kill/adopt by default. Present the landscape and gaps.

此Skill仅用于调研，绝不编写产品代码。
优先使用英文进行广泛搜索，除非用户明确要求使用其他搜索语言。
保存文件时使用用户的语言，除非仓库有更强的约定。
若
```
what
```
、
```
who
```
或
```
why
```
不明确，最多询问2个分类问题，每次一个。
尽可能同时运行所有4个调研通道。
所有结论需有来源支持。引用、排名和非显而易见的结论需包含链接。
去重以不同名称或产品层级出现的工具。
默认不推荐构建/淘汰/采用任何方案，仅呈现格局和差距。

Workflow

工作流

Step 0: Triage

步骤0：分类梳理

Parse the request into:

```
what
```
: the pain point, idea, or capability to survey
```
who
```
: who feels the pain or uses the workflow
```
why
```
: why it matters now

Before proceeding, check whether

.survey/{slug}/triage.md

already exists.

If it exists, ask whether to reuse or overwrite.
If the user keeps the existing survey, skip to Step 3 and summarize the saved files.

Save:

markdown

undefined

将请求解析为：

```
what
```
：需要调研的痛点、想法或能力
```
who
```
：受影响的人群或工作流使用者
```
why
```
：当前开展调研的必要性

开始前，检查

.survey/{slug}/triage.md

是否已存在。

若已存在，询问用户是复用还是覆盖。
若用户选择复用现有调研，直接跳至步骤3并总结已保存的文件。

保存内容：

markdown

undefined

Triage

分类梳理

Problem: {what}
Audience: {who}
Why now: {why}

undefined

问题：{what}
受众：{who}
当前必要性：{why}

undefined

Step 1: Run 4 Parallel Lanes

步骤1：运行4个并行通道

Launch all lanes together.

同时启动所有通道。

Lane A: Context

通道A：背景调研

Research:

where the problem appears in real workflows
who is affected
how people currently work around it
adjacent problems and downstream consequences
direct user voices from communities

Return sections:

```
## Workflow Context
```
```
## Affected Users
```
```
## Current Workarounds
```
```
## Adjacent Problems
```
```
## User Voices
```

研究内容：

问题在实际工作流中的出现场景
受影响人群
人们当前的临时解决方案
关联问题和后续影响
来自社区的直接用户反馈

返回章节：

```
## 工作流背景
```
```
## 受影响用户
```
```
## 当前临时解决方案
```
```
## 关联问题
```
```
## 用户反馈
```

Lane B: Solutions

通道B：解决方案调研

Research:

products, plugins, libraries, SaaS, GitHub projects, services
curated lists and comparison pages
common pricing and packaging
limitations and repeated complaints

Return sections:

```
## Solutions
```
```
## Frequency Ranking
```
```
## Categories
```
```
## Curated Sources
```

研究内容：

产品、插件、库、SaaS、GitHub项目、服务
精选列表和对比页面
常见定价和打包方式
局限性和反复出现的投诉

返回章节：

```
## 解决方案
```
```
## 提及频率排名
```
```
## 分类
```
```
## 精选来源
```

Lane C: Actual Behavior

通道C：实际使用行为调研

Research:

what people actually use in practice
which manual workflows persist despite vendor claims
common frustration patterns
where users drop back to spreadsheets, scripts, copy-paste, or multi-tool workflows

Return sections:

```
## What People Actually Use
```
```
## Common Workarounds
```
```
## Pain Points With Current Solutions
```
```
## Sources
```

研究内容：

人们实际使用的方案
尽管厂商宣称已有解决方案，但仍持续存在的手动工作流
常见的痛点模式
用户回归到电子表格、脚本、复制粘贴或多工具工作流的场景

返回章节：

```
## 实际使用方案
```
```
## 常见临时解决方案
```
```
## 当前方案的痛点
```
```
## 来源
```

Lane D: Alternatives / Platform Map

通道D：替代方案/平台映射

Default mode:

JTBD alternatives
cross-industry substitutes
indirect competitors

For agent / model / orchestration topics, this lane must instead normalize the space into:

```
## Settings
```
```
## Rules
```
```
## Hooks
```
```
## Platform Gaps
```

Use a comparison table for Claude / Codex / Gemini when applicable.

默认模式：

JTBD（Jobs To Be Done）替代方案
跨行业替代品
间接竞品

对于Agent / 模型 / 编排类主题，此通道必须将领域标准化为：

```
## Settings
```
```
## Rules
```
```
## Hooks
```
```
## 平台差距
```

适用时使用对比表格展示Claude / Codex / Gemini的差异。

Step 2: Synthesize and Save

步骤2：合成并保存

Write

context.md

markdown

undefined

编写

context.md

：

markdown

undefined

Context: {project name}

背景：{项目名称}

Workflow Context

工作流背景

{when and where the problem shows up}

{问题出现的时间和场景}

Affected Users

受影响用户

Role	Responsibility	Skill Level
{rows}

角色	职责	技能水平
{行内容}

Current Workarounds

当前临时解决方案

{numbered list with limitations}

{带局限性的编号列表}

Adjacent Problems

关联问题

{lettered or bulleted list}

{字母或项目符号列表}

User Voices

用户反馈

{quotes with source links}


Write `solutions.md`:

```markdown

{带来源链接的引用}


编写`solutions.md`：

```markdown

Solution Landscape: {project name}

解决方案格局：{项目名称}

Solution List

解决方案列表

Name	Approach	Strengths	Weaknesses	Notes
{deduplicated list}

名称	实现方式	优势	劣势	说明
{去重后的列表}

分类

{grouped by category}

{按分类分组的内容}

What People Actually Use

实际使用方案

{behavior summary}

{行为总结}

Frequency Ranking

提及频率排名

{most mentioned solutions}

{被提及最多的解决方案}

Key Gaps

核心差距

{structural gaps not covered today}

{当前未被覆盖的结构性差距}

Contradictions

矛盾点

{marketed claims vs user reality}

{厂商宣传与用户实际体验的差异}

Key Insight

核心洞察

{1 paragraph}


When the topic is platform or agent related, write `platform-map.md`:

```markdown

{1段总结}


当主题涉及平台或Agent时，编写`platform-map.md`：

```markdown

Platform Map: {project name}

平台映射：{项目名称}

Settings

Concern	Claude	Codex	Gemini	Common Layer
{rows}

关注点	Claude	Codex	Gemini	通用层
{行内容}

Rules

Concern	Claude / OMC	Codex / OMX	Gemini / OHMG	Common Layer
{rows}

关注点	Claude / OMC	Codex / OMX	Gemini / OHMG	通用层
{行内容}

Hooks

Lifecycle	Claude	Codex	Gemini	Common Layer
{rows}

生命周期阶段	Claude	Codex	Gemini	通用层
{行内容}

Platform Gaps

平台差距

{where abstractions do not align cleanly}

undefined

{抽象层无法完美对齐的场景}

undefined

Step 3: Summarize for the User

步骤3：为用户生成总结

Return a short summary:

markdown

undefined

返回简短总结：

markdown

undefined

Survey complete: {slug}

调研完成：{slug}

Context

背景

{1-2 sentence summary}
Main workaround: {most common workaround}

{1-2句话总结}
主要临时解决方案：{最常见的方案}

Solution Landscape

解决方案格局

{N} solutions across {M} categories
Key insight: {one sentence}
Key gap: {one sentence}

涵盖{M}个分类的{N}个解决方案
核心洞察：{1句话}
核心差距：{1句话}

Files

生成文件

```
.survey/{slug}/triage.md
```
```
.survey/{slug}/context.md
```
```
.survey/{slug}/solutions.md
```
```
.survey/{slug}/platform-map.md
```
# if created


Do not move into planning or implementation unless the user asks.

---

```
.survey/{slug}/triage.md
```
```
.survey/{slug}/context.md
```
```
.survey/{slug}/solutions.md
```
```
.survey/{slug}/platform-map.md
```
# 若已创建


除非用户要求，否则不要进入规划或实施阶段。

---

Quick Reference

快速参考

Action	Instruction
Start a survey	`survey <topic>`
Reuse existing results	Check `.survey/{slug}/` first
Search strategy	Search in English, write in the user's language
Parallel lanes	Context + Solutions + Behavior + Alternatives/Platform Map
Next step after survey	`/plan` , `jeo` , `ralph` , or implementation if the user asks

操作	说明
启动调研	`survey <主题>`
复用现有结果	先检查 `.survey/{slug}/`
搜索策略	用英文搜索，用用户的语言输出
并行通道	背景+解决方案+实际行为+替代方案/平台映射
调研后下一步	若用户要求，使用 `/plan` 、 `jeo` 、 `ralph` 或开始实施

Final Checklist

最终检查清单

```
what / who / why
```
are clear
Existing survey checked before overwrite
4 lanes executed in parallel
Source links included for quotes and non-obvious claims
```
context.md
```
saved
```
solutions.md
```
saved
```
platform-map.md
```
saved for agent/platform topics
Final user summary is factual, short, and recommendation-free

```
what / who / why
```
已明确
覆盖前已检查现有调研
已并行执行4个通道
引用和非显而易见的结论包含来源链接
已保存
```
context.md
```
已保存
```
solutions.md
```
Agent/平台类主题已保存
```
platform-map.md
```
最终用户总结基于事实、简洁且无推荐倾向

survey

Original

Translation

survey - Cross-Platform Problem-Space Scan

survey - 跨平台问题领域调研

When to use this skill

何时使用此Skill

Do not use this skill when

何时不使用此Skill

Output Package

输出成果包

Core Abstraction Layer

核心抽象层

Platform Adapter

平台适配层

Execution Policy

执行规则

Workflow

工作流

Step 0: Triage

步骤0：分类梳理

Triage

分类梳理

Step 1: Run 4 Parallel Lanes

步骤1：运行4个并行通道

Lane A: Context

通道A：背景调研

Lane B: Solutions

通道B：解决方案调研

Lane C: Actual Behavior

通道C：实际使用行为调研

Lane D: Alternatives / Platform Map

通道D：替代方案/平台映射

Step 2: Synthesize and Save

步骤2：合成并保存

Context: {project name}

背景：{项目名称}

Workflow Context

工作流背景

Affected Users

受影响用户

Current Workarounds

当前临时解决方案

Adjacent Problems

关联问题

User Voices

用户反馈

Solution Landscape: {project name}

解决方案格局：{项目名称}

Solution List

解决方案列表

Categories

分类

What People Actually Use

实际使用方案

Frequency Ranking

提及频率排名

Key Gaps

核心差距

Contradictions

矛盾点

Key Insight

核心洞察

Platform Map: {project name}

平台映射：{项目名称}

Settings

Settings

Rules

Rules

Hooks

Hooks

Platform Gaps

平台差距

Step 3: Summarize for the User

步骤3：为用户生成总结

Survey complete: {slug}

调研完成：{slug}

Context

背景

Solution Landscape