by 安全扫描
OpenClaw
可疑
medium confidenceThe skill's instructions are broadly coherent with a read-only analytics integration, but there are notable mismatches (it asks for and instructs writing an API key to local MCP config files yet declares no required credential or config paths), so proceed with caution.
评估建议
Key points before installing/using this skill:
- The skill will ask for and use an Outlit MCP API key and tells the agent how to add it to local MCP config files (project .cursor/mcp.json, Claude desktop config in your OS profile) or call a 'claude mcp add' command. The metadata does not list any required credential or config paths — verify this discrepancy with the publisher.
- Do not paste high-privilege or long-lived org-wide API keys into chat. If you try it, prefer a limited-scope, read-o...详细分析 ▾
⚠ 用途与能力
The skill purports to provide read-only analytics access to Outlit via MCP tools (outlit_*), which matches the documented tools and queries. However, the SKILL.md explicitly requires an API key and instructs adding an MCP server entry (Authorization: Bearer API_KEY) to local MCP configs/CLI, while the skill metadata declares no primary credential, no required env vars, and no required config paths — an incoherence between declared requirements and actual setup steps.
⚠ 指令范围
Runtime instructions tell the agent to detect the environment and modify local configuration files (project/.cursor/mcp.json, ~/Library/Application Support/Claude/claude_desktop_config.json, %APPDATA%/Claude/...), or run a CLI command (claude mcp add ...). Those are outside of pure query semantics and grant the skill the ability to write to user config locations unless the agent explicitly asks the user first. The instructions also require asking the user for their API key if not provided, which implies handling secrets during runtime.
✓ 安装机制
There is no install spec and no code files to execute — this is instruction-only, which minimizes written-to-disk risk. The skill references an external MCP endpoint (https://mcp.outlit.ai/mcp) but does not instruct downloading arbitrary code.
⚠ 凭证需求
Requesting an Outlit API key is proportional to the stated purpose (service access). However, the skill metadata does not declare any required environment variable or primary credential while the instructions clearly require a secret API_KEY and show writing it into local config files — this mismatch is suspicious and reduces transparency about what secrets will be used/stored.
⚠ 持久化与权限
always:false (good). But the instructions explicitly tell the agent how to persist the API key into various MCP configuration files and to call a CLI to register the server; that means the skill, if followed automatically, could cause persistent credential storage in user config locations. The metadata didn't disclose these config paths, so the skill would gain persistent presence in user config without that being advertised.
安全有层次，运行前请审查代码。
运行时依赖
无特殊依赖
版本
latestv1.0.02026/2/16
Initial release of the Outlit MCP skill for querying customer intelligence data via MCP tools. - Supports six tools: customer and user browsing, deep dives, timelines, custom SQL analytics, and schema discovery. - Detailed parameter and usage guides for each tool (outlit_list_customers, outlit_list_users, outlit_get_customer, outlit_get_timeline, outlit_query, outlit_schema). - Comprehensive setup instructions for API key and MCP integration across multiple environments (Claude Code, Cursor, Claude Desktop). - Includes best practices, gotchas, data model overview, and known limitations. - Emphasizes always discovering schema before writing SQL and filtering at the source for performance.
● 可疑
安装命令
点击复制官方npx clawhub@latest install outlit-mcp
镜像加速npx clawhub@latest install outlit-mcp --registry https://cn.longxiaskill.com