Inspect OpenClaw skill code, network calls, and secret handling before approval.
npx skills add https://github.com/currentsuspect/currentsuspect --skill clawhub-safe
Please help me install this Agent Skill. Skill: clawhub-safe Source: https://github.com/currentsuspect/currentsuspect/tree/main/skills/clawhub-safe Command: npx skills add https://github.com/currentsuspect/currentsuspect --skill clawhub-safe
ClawHub Safe mitigates the risks of prompt-injection when using OpenClaw skills by enabling users to discover and vet skills thoroughly before installation.
Search for a skill, review its details, flag risks, ask for approval, and install it with: 'clawhub install <slug> --force'.
High-intent search queries and answers about installing and using this skill.
To vet skills for prompt-injection risks, manually inspect network requests, code security, and secret handling before activation. This process minimizes vulnerabilities by ensuring only reviewed skills receive installation approval.
Skill discovery is the process of searching and vetting skills to minimize prompt-injection risks before installation. It reduces vulnerabilities by allowing users to manually inspect skill content, network requests, and secret handling to assess potential risks prior to activation.
To securely install an OpenClaw skill after reviewing its code, search for the skill, review its details, flag risks, ask for approval, and execute the installation using the command 'clawhub install <slug> --force' to ensure explicit user consent.
Yes, you can manually inspect network calls and secret handling within a skill library. The skill review process empowers users to review code security and assess potential risks before granting approval for skill installation.
Yes, approval-based installation is needed to manage prompt-injection risks effectively. It ensures skills are installed only after explicit user approval, mitigating threats by forcing a review of network requests and code security beforehand.
ryukyagamilightCapture and analyze network packets using tcpdump and tshark.
MomoDavilukeVet AI agent skills through source checks, code review, and risk classification.
ArianHobson333Automate source code security scanning for vulnerabilities, CVEs, and hardcoded secrets.
wongdeanGenerate Docker sandbox profiles to restrict OpenClaw skill access to files, network, and processes.
