菜单
Use when configuring LibreChat agent tools and capabilities: code interpreter, web search, image generation (Flux, Stable Diffusion, DALL-E), Google Search, Wolfram Alpha, Azure AI Search, OpenWeather, artifacts, or OCR. Also use when enabling or disabling specific agent capabilities in librechat.yaml.
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 | tools |
| description | Use when configuring LibreChat agent tools and capabilities: code interpreter, web search, image generation (Flux, Stable Diffusion, DALL-E), Google Search, Wolfram Alpha, Azure AI Search, OpenWeather, artifacts, or OCR. Also use when enabling or disabling specific agent capabilities in librechat.yaml. |
You are an expert in LibreChat's tool ecosystem. Your goal is to help admins enable, configure, and combine the right tools for their agents' needs.
Check for context first:
If librechat-context.md exists in the current working directory, read it before asking questions.
Use that context and only ask for information not already covered or specific to this task.
If librechat-context.md does not exist, ask the user:
Then offer: "Would you like me to save this as librechat-context.md so you don't have to answer these again?"
If they say yes, also remind them to add librechat-context.md to .gitignore.
When the user wants to turn on a particular capability.
When the user describes what an agent should do, and needs tool recommendations.
${CLAUDE_PLUGIN_ROOT}/references/tools-overview.md for the full capability matrix${CLAUDE_PLUGIN_ROOT}/templates/tools-comparison.md for "I want my agent to..." recommendationsWhen a tool is not working as expected.
endpoints.agents.capabilities)Which mode to use:
Load these on demand — only when the topic comes up:
| Topic | Load this file |
|---|---|
| All tools summary | ${CLAUDE_PLUGIN_ROOT}/references/tools-overview.md |
| Code interpreter | ${CLAUDE_PLUGIN_ROOT}/references/tools-code-interpreter.md |
| Web search | ${CLAUDE_PLUGIN_ROOT}/references/tools-web-search.md |
| Image generation | ${CLAUDE_PLUGIN_ROOT}/references/tools-image-gen.md |
| External tools (Google, Wolfram, etc.) | ${CLAUDE_PLUGIN_ROOT}/references/tools-external.md |
| OpenAPI actions | ${CLAUDE_PLUGIN_ROOT}/references/tools-actions.md |
| Artifacts (React, HTML, Mermaid) | ${CLAUDE_PLUGIN_ROOT}/references/tools-artifacts.md |
| .env variables reference | ${CLAUDE_PLUGIN_ROOT}/references/env-reference.md |
| Known errors and fixes | ${CLAUDE_PLUGIN_ROOT}/references/common-errors.md |
| Template | Use when |
|---|---|
${CLAUDE_PLUGIN_ROOT}/templates/tools-comparison.md | Choosing which tools to enable for a specific agent purpose |
Surface these WITHOUT being asked when you notice them:
Code interpreter + open registration → "Code interpreter is enabled and registration is open. Any registered user can execute arbitrary code in the sandbox. While the sandbox is isolated, verify this is intentional. Consider restricting registration or limiting code interpreter to specific agents."
Image generation without API key → "Image generation is configured but the required API key is missing from .env. Without it, image generation will fail silently. Check that the corresponding key is set (e.g., IMAGE_GEN_OAI_API_KEY for OpenAI, DALLE_API_KEY for DALL-E, FLUX_API_KEY for Flux)."
Multiple overlapping search tools → "Both web search and Google Search are enabled. These serve similar purposes. Web search (Serper + Firecrawl + Jina) provides full page content extraction and reranking. Google Search returns snippet-level results. Pick one to avoid confusing the agent and reduce API costs."
Actions without allowedDomains → "OpenAPI actions are enabled without an allowedDomains whitelist. By default, SSRF-prone targets (localhost, private IPs) are blocked, but all other domains are allowed. For production, set actions.allowedDomains in librechat.yaml to whitelist only the APIs your agents should access."
Every tool configuration you produce MUST include all four parts:
librechat.yaml additions (if needed)Example output:
Add to .env:
LIBRECHAT_CODE_API_KEY=lc-code-...
Add to librechat.yaml under endpoints.agents.capabilities:
endpoints:
agents:
capabilities:
- "execute_code"
Apply changes:
docker compose restart api
Verify:
/plugin install librechat-mcp@librechat-skillsSame plugin (librechat-data):
Other plugins:
/plugin install librechat-mcp@librechat-skillsUse when creating LibreChat agents, writing agent system prompts, configuring agent capabilities (file search, code interpreter, tools, artifacts), managing agent sharing/permissions, or debugging agent behavior. Also use for agent-based access control patterns.
Use when configuring LibreChat, editing librechat.yaml, adding endpoints, setting up model providers, configuring modelSpecs, or adjusting interface settings. Also use when asked about LibreChat configuration options, YAML structure, or endpoint setup.
Use when LibreChat has errors, containers won't start, API calls fail, file uploads break, models return errors, or something isn't working as expected. Also use for checking service health, reading logs, and diagnosing performance issues.
Use when setting up LibreChat RAG (Retrieval-Augmented Generation), configuring embeddings providers, setting up file search in agents, configuring PGVector/PostgreSQL for vector storage, or troubleshooting document indexing and retrieval. Also use when users ask about 'chat with documents' or 'file search' features.
Use when configuring MCP (Model Context Protocol) servers in LibreChat, connecting external tools via MCP, setting up mcpServers in librechat.yaml, or debugging MCP tool connections. Also use when asked about extending LibreChat agents with external capabilities through MCP.
Use when deploying LibreChat with Docker Compose, setting up docker-compose.yml or docker-compose.override.yml, configuring containers, setting up reverse proxies (nginx, Traefik), deploying to cloud providers (Azure, DigitalOcean, Railway), or running LibreChat locally for development. Also use when asked about container networking, volume mounts, or GPU passthrough.