This scanner runs as an MCP server to expose security analysis tools directly to Claude and other AI assistants. It provides four core tools: prism_scan analyzes code for shell injection, data exfiltration, persistence mechanisms, and 38 other behavioral patterns across skills and plugins. prism_grade assigns A-F scores based on severity. prism_clean_scan checks your system for leftover artifacts like LaunchAgents and crontab entries. prism_clean_plan generates removal plans without executing them. Reach for this when you're evaluating third-party agent code before installation or need to audit what previous integrations left behind on your machine. It works locally with GitHub repos or filesystem paths.
claude mcp add --transport stdio aidongise-cell-prism-scanner uvx prism-scanner