菜单
Look things up on Wikipedia with the oc CLI, which renders pages as compact text with numbered actions and ships a tuned Wikipedia shortcut. Use it to search Wikipedia, read articles, and read non English wikis in minimal tokens.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/only-cli/benchmarks --skill browse-oc-wiki命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Look things up in language documentation with the oc CLI, which renders pages as compact text with numbered actions and ships tuned shortcuts for the major language references. Use it to search and read Python, MDN, Node.js, Ruby, Go, Rust, Java, PHP, TypeScript, C/C++, and .NET docs in minimal tokens.
Read web pages with the built in WebFetch tool, which fetches a URL and returns its content processed by a model against your prompt. Use it to read Wikipedia articles and search result pages.
Find facts with the built in WebSearch tool, which runs a web search and returns result snippets. Use it to answer questions from Wikipedia content without fetching pages yourself.
正在显示 SKILL.md
| name | browse-oc-wiki |
| description | Look things up on Wikipedia with the oc CLI, which renders pages as compact text with numbered actions and ships a tuned Wikipedia shortcut. Use it to search Wikipedia, read articles, and read non English wikis in minimal tokens. |
| allowed-tools | Bash(oc:*) |
Wikipedia shortcuts, which resolve to a URL and then render like any other page:
oc wiki search <query> searches Wikipedia and lists the matching articles, each with a snippet, its size, and when it was last edited.oc wiki article <title> renders that article. The title takes every word after it, so oc wiki article Eiffel Tower works unquoted.oc wiki lang <code> <title> renders the article from another language wiki: oc wiki lang de Berlin reads the German article.The generic commands work on whatever is open:
oc do <n> follows the link behind number [n] on the page you just rendered. Following a search result is always oc do <n>, never a second command with a URL.oc find <query> prints every place a string appears on the page already open, one line per match with the number to read it by.oc read <n> prints the region at [n] in full: one block with its context, or a whole section when [n] is a heading.oc next prints the next ~500 tokens of the page already open, continuing where the last view stopped.oc raw [url] returns the whole page as distilled markdown. With no URL it uses the page already open.oc open <url> renders any URL, for when you have one already.Tips:
... 164 more blocks (~7,100 tokens) at the end is the price of the rest of the page.oc find <query> then oc read <n> on the match is the cheapest way to one fact. Reach for oc raw only when you truly need the whole page.& or ?.