Analyze AI agent configurations for secrets, PII, prompt injection, and backdoors.
npx skills add https://github.com/XiaoYiWeio/deepsafe-scan --skill deepsafe-scan
Please help me install this Agent Skill. Skill: deepsafe-scan Source: https://github.com/XiaoYiWeio/deepsafe-scan/tree/main Command: npx skills add https://github.com/XiaoYiWeio/deepsafe-scan --skill deepsafe-scan
DeepSafe Scan provides a preflight security check for AI coding agents by examining deployment/posture configs, installed Skills and MCP servers, memory/sessions, and agent hook configurations for secrets, PII, prompt injections, and backdoors, enabling safer AI workflows.
Run the scanner on your project to generate a comprehensive security report.
High-intent search queries and answers about installing and using this skill.
To scan AI agent configurations for prompt injection and backdoors, you run a static analysis check on deployment setups and installed MCP servers. This process examines memory, hooks, and skills to detect dangerous calls and exfiltration patterns.
The best way to detect secrets and PII leakage in AI agent memory is to perform a dedicated memory and session review. This static analysis identifies exposed credentials and personal data while assessing prompt injection risks.
Yes, the security scanner works with Claude Code, Cursor, Codex, OpenClaw, and OpenCode workflows. It applies cross-platform compatibility to analyze posture, memory, and hooks configurations to identify credential exposure.
Yes, you need Python3 installed to run static analysis on MCP servers and agent hooks. The scanner uses Python to examine directories for hardcoded secrets, dangerous calls, and data exfiltration patterns.
Yes, you can use optional LLM-enhanced analysis to detect complex attack patterns in agent skills. This semantic analysis supplements static scanning to surface sophisticated exfiltration risks and backdoor vectors.
Posture analysis may show false positives when static analysis flags benign configurations as misconfigurations. Applying optional LLM-assisted semantic analysis helps contextualize the findings and reduce false positives.
xlelibertyScans AI agent skill directories for security vulnerabilities via static analysis.
kardelitaituAnalyze repository files for secrets, vulnerabilities, PII, and configuration risks.
tharunramagiriIntegrate Gmail MCP tools into containerized AI agents with OneCLI-managed OAuth credentials.
TECHKNOWMAD-LABSScan code repositories for vulnerabilities and leaked secrets using static analysis tools.
