Web Intel

网络搜索与信息提取统一入口。整合 content-extraction、Jina Reader、Firecrawl、web-access CDP。自动选最经济工具链。可被 BMW（brand-marketing-workflow）、wealth 子代理等直接调用。

This skill appears coherent with its stated role as a router between local tools and r.jina.ai. Before enabling it: (1) Confirm you trust the installed firecrawl binary and understand what network calls it performs, since the script invokes it via subprocess. (2) Be aware that deep mode will use a local CDP proxy (localhost:3456) and therefore can read pages that require your browser's login/session state — only enable deep-mode usage if you trust the web-access CDP proxy and the skill. (3) Verify the presence and integrity of the referenced skills/content-extraction scripts in your workspace; if those are missing the skill will degrade but still run other backends. (4) If you have strong privacy requirements, test the script in a sandboxed environment first to observe network calls and outputs.

Type: OpenClaw Skill Name: web-intel Version: 1.0.0 The web-intel skill provides a routing layer for web search and content extraction, utilizing Firecrawl, Jina Reader, and a local CDP proxy. While the Python implementation in `scripts/web_intel.py` follows safe coding practices (e.g., using list-based arguments for `subprocess.run` to prevent shell injection), the skill possesses high-risk capabilities including shell execution, external network access to `r.jina.ai`, and direct interaction with the user's browser session (including login states) via a CDP proxy at `localhost:3456`. These features are aligned with the stated purpose in `SKILL.md` but represent a significant attack surface and high-privilege access, warranting a suspicious classification under the provided criteria for risky capabilities.