Extract content from a GitHub url.

This skill appears to do what it says — fetch content from public GitHub URLs — but check these points before installing: (1) SKILL.md and the script call 'wget' though the registry metadata only lists 'uv'; ensure 'wget' is available on the host or adjust expectations. (2) The script declares Python dependencies (furl, requests, loguru, urllib3) but there is no install specification — confirm how your runtime will provide or install them. (3) The tool will download arbitrary user-supplied URLs (via requests/wget) and write them to a temporary directory — only use it with trusted GitHub URLs. (4) No secrets are requested and the code only talks to GitHub/raw.githubusercontent.com, but if you run this in a shared environment be aware it will perform outbound HTTP requests and write files to /tmp. If any of these inconsistencies are unacceptable, ask the author to: add 'wget' to required binaries, document dependency installation in the registry, or provide an explicit install spec so you can review what will be installed.

Type: OpenClaw Skill Name: gh-extract Version: 0.0.2 The skill is suspicious due to the direct execution of `wget` with a user-provided URL in `gh_extract.py`. The `SKILL.md` instructs the AI agent to pass user input directly as the `<url>` argument to the Python script, which then uses `subprocess.run(['wget', ..., url])`. While `subprocess.run` with a list mitigates direct shell injection, it exposes the `wget` command to potential vulnerabilities if a crafted URL can exploit `wget` itself (e.g., arbitrary file writes or SSRF), representing a lack of robust input sanitization for external command execution.