Agent Skills › ruvnet/open-claude-code › occ-use

occ-use

GitHub

指导使用 Open Claude Code CLI 执行编码任务。涵盖安装验证、API Key 配置及受限的读写操作，强调安全规范与结果反馈。

skills/occ-use/SKILL.md ruvnet/open-claude-code

触发场景

要求使用 OCC 进行编码选择 OCC 提供商通过 CLI 执行受限编码任务

安装

npx skills add ruvnet/open-claude-code --skill occ-use -g -y

更多选项

不安装直接使用

npx skills use ruvnet/open-claude-code@occ-use

指定 Agent (Claude Code)

npx skills add ruvnet/open-claude-code --skill occ-use -a claude-code -g -y

安装 repo 全部 skill

npx skills add ruvnet/open-claude-code --all -g -y

预览 repo 内 skill

npx skills add ruvnet/open-claude-code --list

SKILL.md

Frontmatter

{
    "name": "occ-use",
    "description": "Run Open Claude Code (OCC) for coding tasks and verify its CLI installation. Use when asked to use OCC, select its provider, or perform a bounded coding task through its CLI."
}

Use Open Claude Code

Identify the requested repository and task. Inspect its instructions before executing.
Check the pinned CLI without a provider call: npx --yes @ruvnet/open-claude-code@2.1.3 --version Expect v2.1.3. Use --help for the exact supported flags.
Use an existing provider binding. Anthropic uses ANTHROPIC_API_KEY; OpenAI compatible providers use OPENAI_API_KEY and optionally OPENAI_BASE_URL. Never display, copy, or persist credentials. This skill supplies no credentials or permissions.
For a read task, restrict tools and turns: npx --yes @ruvnet/open-claude-code@2.1.3 --print --allowedTools Read --max-turns 4 "Read README.md and summarize the project." For OpenAI compatible providers, also pass --model gpt-4.1-mini (or another supported OpenAI model ID); routing depends on the model name, not the API key alone. Provider calls may incur charges.
For edits, preserve repository instructions and user scope. Do not assume that installing this skill authorizes publishing, deletion, or external messages.
Report the command exit status, observed result, and any missing provider access. Do not claim full upstream Claude Code parity.

If asked only to explain or plan, do not execute the task. If asked to verify installation, stop after version and help unless live inference was requested.