| name | snow-docs |
| description | Use when the user asks to install, configure, troubleshoot, or understand Snow CLI itself — Profile/API setup, MCP servers, Skills, Hooks, sub-agents, sensitive commands, proxy/browser, third-party relay, LSP/ACE, Team mode, SSE, privacy, plugins, or docs paths. Loads official bundled usage docs via progressive disclosure (list/search/get) instead of guessing from model memory. |
| allowed-tools | snow-docs-list, snow-docs-search, snow-docs-get, filesystem-read, askuser-ask_question |
snow-docs
Authoritative Snow CLI usage documentation skill (bundled with the CLI version).
When to use
Trigger for any request about Snow itself, for example:
- First-time setup / Profile / API / model selection
- MCP install/config/enable/disable / troubleshooting
- Skills (
~/.snow/skills, .snow/skills, skill-execute)
- Hooks, sub-agents, sensitive commands, YOLO
- Proxy / browser / third-party relay / custom headers
- LSP, Team mode, SSE, StatusLine, privacy, plugins, games
Do not use this skill for general coding in an unrelated project unless the user is configuring Snow.
Progressive disclosure workflow
- Discover with
snow-docs-list (catalogue only) or snow-docs-search (keywords).
- Read one topic with
snow-docs-get using an id from search/list (e.g. 14.MCP配置.md).
- Then inspect/edit local config files only as the docs specify.
- Prefer bundled docs over model memory or random GitHub raw URLs.
Never load all documents in one turn. Never dump the entire manual into context.
Authoritative config locations (confirm via docs)
Common paths (always re-check the fetched doc for the current version):
~/.snowcli/ — global CLI config / profiles
~/.snow/ — global skills, todos, etc.
<project>/.snow/settings.json — project settings (disabledSkills, disabledBuiltInServices, ...)
<project>/.snow/skills / ~/.snow/skills — user skills
Safety rules
- Tools are read-only for docs. Writing config still goes through normal file/UI flows.
- Confirm high-risk changes with the user (API keys, disabling security, sensitive command rules).
- Do not silently rewrite global config.
- Disable path: users can disable the
snow-docs skill (disabledSkills) or the snow-docs built-in service (disabledBuiltInServices).
Locale
Docs tools follow the user Language Settings (zh / zh-TW → Chinese docs, otherwise English). Pass locale: "zh" | "en" only when the user asks for a specific language.
Quick map
| Topic | Search hints |
|---|
| Install / update | install, 安装 |
| First config / Profile | profile, 首次配置, api |
| MCP | mcp, tools |
| Skills | skills, skill-execute |
| Hooks | hooks |
| Sub-agents | subagent, 子代理 |
| Sensitive commands | sensitive, 敏感命令 |
| Relay / headers | relay, 中转 |
| LSP / ACE | lsp, ace |
| Team / SSE / privacy | team, sse, privacy |
After loading this skill, start with snow-docs-search or snow-docs-list for the user's topic.