Smart Skill Finder

What to check before installing/using this skill: 1) Review and sanitize inputs: the Skills-CLI search runs via subprocess.run with shell=True using user-derived text. If you or the agent runs this code on arbitrary user input, command-injection is possible. Prefer escaping or using a list form for subprocess or otherwise sanitizing the query string. 2) Verify claimed features: the README/SKILL.md advertise semantic-model based understanding, VirusTotal checks, and workspace/memory integration, but the code uses simple keyword matching and I found no VirusTotal or local workspace reads. Treat those claims as marketing until the author provides evidence or code updates. 3) Be cautious with browser automation: the browser discovery code assumes a browser_controller and will drive a browser (open pages, type, press keys). Only run this in a controlled environment and avoid passing credentials or sensitive context to it. 4) Installation commands are returned (git clone, npx, clawhub). Do not execute them automatically — inspect sources (GitHub repo, author) before cloning/running third-party skills. Consider cloning into a sandboxed VM/container and scanning the repo before use. 5) If you plan to run this skill in production or allow it to be invoked autonomously, request the author to: - Remove shell=True usage or properly escape/validate user inputs; use subprocess.run([...]) list form. - Make security-scanning calls explicit (show how VirusTotal/OpenClaw are queried and where API keys would be stored if needed). - Document the browser_controller contract and provide safe default behavior. Given these issues (misleading claims + a real shell-injection pattern), review and remediation are recommended before trusting the skill in an environment with sensitive data or in which the agent can run code autonomously.