Scan AI agent skills and inputs for injection and exfiltration threats.
npx skills add https://github.com/gula00/autoclaw-skills --skill clawdefender-gula00
Please help me install this Agent Skill. Skill: clawdefender Source: https://github.com/gula00/autoclaw-skills/tree/main/clawdefender-1 Command: npx skills add https://github.com/gula00/autoclaw-skills --skill clawdefender-gula00
This Skill acts as a robust security layer for AI agents, proactively identifying and mitigating risks associated with malicious code, prompt injection, and data exfiltration.
Scan all installed skills for security issues using the command: ./scripts/clawdefender.sh --audit
High-intent search queries and answers about installing and using this skill.
To protect against prompt injection and command injection, you need a security layer that sanitizes external inputs and scans skill code for malicious patterns. This prevents attackers from executing unauthorized commands through external text or API responses.
SSRF prevention blocks unauthorized server requests by validating URLs before fetching. Checking URLs prevents data exfiltration attacks by ensuring your AI agent does not access internal network resources or send sensitive data to malicious external endpoints.
You can audit installed skills for security vulnerabilities by running a script that analyzes skill code for malicious patterns. This security auditing process detects known threats and mitigates risks associated with malicious code before execution.
Yes, you can sanitize external API responses by running them through an input sanitization service. This cleans the input to prevent prompt injection and command execution, ensuring no malicious commands are present before processing.
Path traversal detection works during security scanning by utilizing pattern matching to identify attempts to access unauthorized files. The scanner analyzes skill code and external inputs to mitigate these attacks and secure your AI agent.
dexhunterScan installed skills and sanitize external input against injection attacks.
winsorllcScan code repositories for security vulnerabilities and exposed secrets using predefined patterns.
nhonvoDetect workflow state and generate adaptive help for planning, coding, debugging, and deploying.
devendrapratapsinghScan application code for vulnerabilities and hardcoded secrets.
