Skip to main content

snow-docs

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.

Informations de source

Dépôt
MayDay-wpf/snow-cli
Dernière activité de la source
15 juillet 2026 à 02:39
Langue détectée de SKILL.md
anglais
Étoiles
1 069
Forks
88

Options d'installation

Le prompt qui vérifie d'abord la source est sélectionné par défaut. Vous pouvez passer à une commande directe ou télécharger une copie locale.

Vérifiez les fichiers source

Lisez SKILL.md et les fichiers associés affichés par SkillsMP avant de décider de l'installer.

Explorateur de fichiers
2 fichiers

Affichage de SKILL.md

SKILL.md
Instructions source · Aperçu en lecture seule
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 1. **Discover** with `snow-docs-list` (catalogue only) or `snow-docs-search` (keywords). 2. **Read one topic** with `snow-docs-get` using an id from search/list (e.g. `14.MCP配置.md`). 3. **Then** inspect/edit local config files only as the docs specify. 4. 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.
Voir sur GitHub