一键导入
ctx7-skilld
ALWAYS use when writing code importing "ctx7". Consult for debugging, best practices, or modifying ctx7, context7.
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。
菜单
ALWAYS use when writing code importing "ctx7". Consult for debugging, best practices, or modifying ctx7, context7.
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。
基于 SOC 职业分类
| name | ctx7-skilld |
| description | ALWAYS use when writing code importing "ctx7". Consult for debugging, best practices, or modifying ctx7, context7. |
| metadata | {"version":"0.5.4","generated_by":"Anthropic · Haiku 4.5","generated_at":"2026-07-06T00:00:00.000Z"} |
ctx7@0.5.4Tags: canary: 0.2.4-canary.0, latest: 0.5.4
References: package.json • README • Docs • Issues • Discussions • Releases
Use skilld search "query" -p ctx7 instead of grepping .skilld/ directories. Run skilld search --guide -p ctx7 for full syntax, filters, and operators.
This section documents version-specific API changes for ctx7 — prioritize recent major/minor releases when integrating with different versions.
BREAKING: ctx7 login --device flag removed — v0.5.1 changed to always use device-code flow (RFC 8628); the flag is no longer accepted or needed source
BREAKING: ctx7 setup --api-key no longer accepts --universal flag — v0.4.5 silently ignored this flag, creating silent breakage when users expected agent-agnostic setup; the flag is now rejected source
BREAKING: CLI config/credential file locations moved to XDG Base Directory — v0.5.2 migrates ~/.context7/credentials to $XDG_CONFIG_HOME/context7 (default ~/.config/context7), updater state to $XDG_STATE_HOME, and cache to $XDG_CACHE_HOME; migration is best-effort and falls back to legacy path on error source
BREAKING: query-docs MCP tool researchMode parameter removed entirely — v0.4.0 added research mode, but v0.4.1 removed it completely; code passing researchMode will fail silently or error source
BREAKING: CLI ctx7 docs --research flag removed entirely — v0.4.0 added research mode flag, but v0.4.1 removed it completely; passing --research now raises an unknown-flag error source
NEW: OAuth 2.0 device authorization flow (RFC 8628) for ctx7 login and ctx7 setup — v0.5.0 added headless-friendly device-code flow that works on SSH, Codespaces, Docker, and CI; automatically selected when SSH_CONNECTION is set or $DISPLAY is missing; can be forced with ctx7 login --device (until v0.5.1 when it became default) source
NEW: ctx7 upgrade command — v0.4.0 added CLI self-update command with version checks, cached state, and safer guidance for npm/pnpm/bun installs; shows non-blocking notices before interactive commands source
NEW: ctx7 remove command — v0.4.0 added as cleanup counterpart to ctx7 setup, with safer detection and removal behaviour; prompts only for agents with actual Context7 artifacts and preserves non-Context7 MCP config when removing entries source
NEW: ctx7 setup --stdio flag — v0.4.3 added support for configuring Context7 as a local stdio MCP server source
DEPRECATED: Skill commands (install/remove/list) — v0.4.5 deprecated in favour of agent-specific management; ctx7 skills commands now emit a deprecation warning source
CHANGED: ctx7 setup --api-key now prompts for mode choice — v0.4.5 changed to ask whether to set up MCP server or CLI + Skills modes when --api-key is passed without explicit --cli, --mcp, --stdio, --oauth, or -y flags; previously short-circuited to MCP mode only source
CHANGED: ctx7 setup now properly supports --antigravity flag — v0.4.5 fixed Antigravity-specific setup paths (installs to .agent/skills, writes MCP config to ~/.gemini/config/mcp_config.json) and removed the --universal flag which was advertised but non-functional source
NEW: --json flag for ctx7 skills list — v0.4.4 added machine-parseable output emitting { skills: [{ name, path, source }] } where source indicates the agent type (universal, claude, cursor, antigravity); matches existing --json pattern on ctx7 library and ctx7 docs source
FIXED: Git Bash Windows path mangling in ctx7 — v0.5.2 now detects and undoes Git Bash rewriting of leading-slash arguments (e.g. /facebook/react → C:/Program Files/Git/facebook/react), accepts the //owner/repo escape sequence, and points users at workarounds for install layouts it cannot auto-detect source
Also changed: ctx7 docs query support for specific versions · XDG_* env var overrides for config/state/cache · CLAUDE_CONFIG_DIR env var respected for Claude Code global config resolution · Malformed MCP config files handled gracefully during removal detection · @inquirer/core added as direct CLI dependency
Add a rule to your MCP client configuration to automatically invoke Context7 without typing use context7 explicitly — this reduces friction when asking library-related questions and ensures documentation is fetched consistently across your workflow. source
Use the library ID directly in your prompt when you already know the exact library — this skips the library-matching step and retrieves documentation faster (e.g. use context7 with /facebook/react for hooks documentation). source
Specify the library version in your prompt or use the /owner/repo@version syntax to get documentation for a specific release — Context7 automatically matches the appropriate version without manual library ID construction. source
Be specific and detailed with your queries when searching — natural language questions like "How do I implement authentication with middleware?" produce more relevant results than vague keywords like "auth". source
Cache API responses for several hours or days since documentation updates are relatively infrequent — this reduces API calls, improves performance, and leaves more budget for other requests. source
Implement exponential backoff when handling rate limit errors (HTTP 429) — use the Retry-After header value to determine sleep time, with fallback to 2^attempt seconds to avoid thundering herd patterns. source
Get an API key from the Context7 dashboard and use it in all requests — anonymous rate limits are significantly lower than authenticated requests, and API keys are required for higher-tier plans and private library access. source
export CONTEXT7_API_KEY="your-api-key"
Use the CONTEXT7_API_KEY environment variable when initializing the TypeScript SDK — the Context7 client reads this automatically without requiring explicit configuration. source
Wrap TypeScript SDK calls in try-catch blocks and check for Context7Error instances — this distinguishes API errors from unexpected runtime errors and enables graceful fallback handling. source
Sort search results by benchmarkScore when multiple libraries match — the benchmark score indicates documentation quality and reduces the chance of selecting a lesser-documented alternative. source
Use the type: "txt" option with getContext() to receive documentation as plain text — this format is more suitable for passing directly to LLM prompts compared to JSON objects. source
Add a context7.json file at your repository root to control how Context7 parses documentation — use excludeFolders patterns to skip outdated, archived, or non-technical content (e.g. docs/archive, **/deprecated, i18n/*). source
Enable the $schema field in context7.json to get IDE validation and autocomplete — this catches configuration errors early and makes the file self-documenting. source
{
"$schema": "https://context7.com/schema/context7.json",
"projectTitle": "Your Library",
"excludeFolders": ["docs/archive", "**/deprecated"]
}
docs-researcher agent in Claude Code when working on long-running tasks to keep your main context clean — spawn it for focused documentation lookups that don't clutter your conversation thread. sourceALWAYS use when writing code importing "daisyui". Consult for debugging, best practices, or modifying daisyui.
ALWAYS use when writing code importing "@storybloq/storybloq". Consult for debugging, best practices, or modifying @storybloq/storybloq, storybloq/storybloq, storybloq storybloq, storybloq.
Use this skill to generate well-branded interfaces and assets for syn.horse — a fictional cyberpunk/glitch-vaporwave shitpost-friendly product. Contains essential design guidelines, colors, type, fonts, assets, and UI kit components for prototyping.
ALWAYS use when writing code importing "@anthropic-ai/claude-code". Consult for debugging, best practices, or modifying @anthropic-ai/claude-code, anthropic-ai/claude-code, anthropic-ai claude-code, anthropic ai claude code, claude-code-2.1.88, claude code 2.1.88.
ALWAYS use when writing code importing "@nuxtjs/seo". Consult for debugging, best practices, or modifying @nuxtjs/seo, nuxtjs/seo, nuxtjs seo, nuxt-seo, nuxt seo.
ALWAYS use when writing code importing "tsx". Consult for debugging, best practices, or modifying tsx.