
Scanner that checks agent skills for prompt injection, exfiltration and other risks
- Open source
- Open-source agent
- 26.7K GitHub stars
- Security

Scanner that checks agent skills for prompt injection, exfiltration and other risks

Adds policy checks, intent screening, trust scores and audit logs to AI agents

Audits GitHub Actions workflows that run AI coding agents for prompt injection risks
Local input firewall that scans text, files and MCP metadata for prompt injection
Runs locallyFlags fake engagement and prompt injection in AI-agent job board listings
Runs locallyPrompt-injection scanner and tool-call gate for AI agents, run locally over stdio
Runs locallyStatic security scanner for MCP servers, agent skills and npm or PyPI packages
Runs locally
Let an agent pay on NEAR through a vault whose rules a smart contract enforces
Runs locally
Self-hosted fleet of persistent, logged-in stealth Chrome sessions for AI agents
Runs locally
Scan prompts, tool definitions and model output for injection, and guard tool calls
Runs locallySigned, sealed handoffs between AI agents, with trust decided by the receiver
Runs locallyEmail MCP server for Gmail, Outlook and IMAP that fences messages against prompt injection
Runs locallySecrets manager MCP server that runs commands with API keys the model never sees
Runs locallyAuthorization proxy that needs a rule check and AI judges to approve each tool call
Runs locallyScore AI agent runs for quality, PII leaks, prompt injection and cost, on your own machine
Runs locallyTyped, policy-checked server operations for AI agents, with approvals and an audit log
Runs locallyOffline security inspection for MCP configs, metadata and dependencies, run from the CLI
Runs locallyPins MCP tool definitions in a lockfile and flags rug-pulls, poisoning and leaked secrets
Runs locallyRead-only MCP server that lints and budgets CLAUDE.md, AGENTS.md and other steering files
Runs locallyFind and strip zero-width, bidi and tag characters hidden in text, offline
Runs locally