Doc-First Research

Automatically consult official documentation first to accurately resolve configuration, usage, permission, or error issues with tools and APIs.

Security Scan

The skill is an instruction-only

Purpose & Capability

The name/description (consult official docs first) matches the instructions. The SKILL.md explicitly says to fetch official documentation and record URLs/versions; no unrelated credentials, binaries, or installs are requested.

Instruction Scope

Instructions require using a web_fetch-like tool to retrieve official docs and to consult GitHub Issues/Stack Overflow when needed. This is appropriate for the stated purpose, but it does entail network fetching of external web pages (the skill instructs the agent to fetch many official documentation URLs).

Install Mechanism

No install spec and no code files — instruction-only skill with no downloads or installs. Lowest-risk category for install behavior.

Credentials

The skill requests no environment variables, credentials, or config paths. All declared requirements are proportional to its function (none).

Persistence & Privilege

always is false, autonomous invocation is allowed (platform default). The skill does not request permanent presence or modify other skills' config.

Assessment

This skill is internally coherent and low-risk because it only contains written guidance to consult official docs and record source URLs/versions. Things to consider before installing: (1) The runtime instructions require the agent to fetch external web pages — confirm you are comfortable with the agent making outbound HTTP requests and how fetched content is logged/cached by the platform. (2) The skill may direct the agent to consult API docs that could lead to API calls; ensure the agent will not automatically submit any of your credentials or secrets while following the guidance. (3) Confirm the platform's web_fetch/network policy (allowed domains, privacy, retention). If those are acceptable, this skill appears safe and appropriate for its stated purpose.

doc-first

任何外部工具、服务、API 的配置或问题，第一时间查官方文档，力求精准解决。

触发条件

遇到以下情况时自动触发：

• 不确定某个工具/服务的正确配置格式
• 不确定某个命令的用法或选项
• 遇到权限、授权、认证相关问题
• 遇到错误不知道原因
• 任何不确定的操作
• 配置 openclaw.json / 环境变量 / CLI 参数时

执行原则

第一原则：查官方文档

• 优先查官方文档/官方 API 文档
• 用 web_fetch 抓取官方页面
• 英文文档为主，中文文档用于辅助确认（中文翻译常有滞后或不完整）
• 查完官方文档仍无解 → 按序尝试：GitHub Issue → Stack Overflow → 开发者社区搜索
• 每次查文档须说明依据：告诉用户查了哪篇文档、哪个版本（URL + 日期或版本号），方便用户核实

第二原则：精准解决

• 不确定答案时，宁可说"我不确定"也不瞎猜
• 遇到奇怪报错，先查文档再尝试修复
• 配置文件动手前，先确认正确格式和路径

第三原则：如实告知

• 官方文档没有 → 直接说明，不编造
• 找不到解决方案 → 说清楚，而不是给可能错误的答案
• 不确定的地方 → 提出最可能的猜测，但标注不确定

第四原则：验证步骤

改了配置之后，重新对比官方文档，确认格式/路径/选项都匹配，再继续下一步。不匹配就修正，不确定就停下来如实告知用户。

第五原则：版本意识

查文档时先确认版本号或服务版本，尤其是 Discord API 这类有大版本变化的（v10 → v14 格式完全不同）。不确定用哪个版本时，优先用最新版文档，并在回答中注明版本依据。

常见误区

• 以为配置格式和旧版本一样（实际上经常变）
• 记错选项名（比如 --force 写成 --forced）
• 混用不同 API 版本（v1 和 v2 格式完全不同）
• 用错误的路径（比如把 user 配置放进 system 配置）

常用文档入口

服务	文档地址
OpenClaw	https://docs.openclaw.ai/llms.txt
OpenAI	https://platform.openai.com/docs
Anthropic	https://docs.anthropic.com/
GLM (智谱)	https://open.bigmodel.cn/dev/api
MiniMax	https://www.minimaxi.com/document/Guide
阿里云通义千问 (DashScope)	https://qianwen-api.aliyun.com/
百度文心一言 (千帆)	https://qianfan.cloud.baidu.com/
讯飞星火认知大模型	https://www.xfyun.cn/doc/spark/
月之暗面 Kimi (Moonshot)	https://platform.moonshot.cn/
昆仑万维天工AI	https://www.tiangong.cn/
字节扣子 (Coze)	https://www.coze.cn/docs/
微信小程序	https://developers.weixin.qq.com/miniprogram/dev/framework/
Telegram Bot API	https://core.telegram.org/bots/api
DeepSeek	https://api.deepseek.com/docs
Slack API	https://api.slack.com/
Discord API	https://discord.com/developers/docs
GitHub API	https://docs.github.com/rest
Notion API	https://developers.notion.com/docs/getting-started
GMAIL API	https://developers.google.com/gmail/api

规则

不查官方文档不开口。找不到就直说，不编答案。