Security Guardian for
AI Agents
Protect your AI agents from malicious skills, prompt injection, and data exfiltration in real-time.
Send this to your AI agent to get started:
Read https://agentguard.gopluslabs.io/skill.md and follow the instructions to set up AgentGuard
Run the command above to install AgentGuard
Your agent scans every skill automatically before install
Pay $0.001/scan via x402 — no API key needed
The AI Agent Skill Crisis
Public skill registries are the new npm -- and they carry the same risks. Malicious actors are publishing poisoned skills that steal credentials, exfiltrate data, and hijack agent behavior at scale.
One in five skills on public registries contain malicious or dangerous patterns
Hardcoded API keys, tokens, and secrets found in published skill packages
Agent instances running unscanned skills with known vulnerabilities
Critical vulnerabilities in popular skill frameworks actively exploited in the wild
Six-Layer Threat Detection
Comprehensive security analysis purpose-built for the AI agent ecosystem. Every skill is scanned across six independent detection engines.
Credential Leak Detection
160+ patterns for API keys, tokens, private keys, database connection strings, and embedded secrets across all major providers.
Prompt Injection Detection
75+ injection patterns including base64-encoded payloads, zero-width character obfuscation, role hijacking, and system prompt extraction.
Malicious Command Detection
112+ patterns for remote code execution, reverse shells, encoded payloads, pastebin piping, and supply chain compromise techniques.
Data Exfiltration Detection
Detects sensitive path access, HTTP/DNS tunneling, credential harvesting, clipboard scraping, and covert data transfer channels.
Permission Abuse Analysis
Compares declared vs. needed tool permissions, detects dangerous combinations, privilege escalation, and calculates risk scores.
URL Analysis
Malicious domain detection, suspicious TLD analysis, phishing pattern matching, URL shortener tracking, and homograph attack detection.
Three Ways to Protect
Integrate AgentGuard wherever skills are installed, published, or reviewed. One scanner, multiple deployment targets.
Scan API
Integrate directly into your registry or platform. A single POST request returns a full threat analysis with risk scoring.
curl -X POST https://agentguard.gopluslabs.io/v1/scan \
-H "X-API-Key: ag_live_xxx" \
-d '{"skill_code": "...", "platform": "mcp"}'Claude Code Plugin
Scan skills locally before installation. Get real-time threat alerts in your development workflow.
# Install the plugin claude plugin add agentguard # Scan before installing any skill claude skill scan ./my-skill --strict
GitHub Action
Auto-scan skill PRs in CI/CD. Block merges that introduce credential leaks or malicious patterns.
# .github/workflows/scan.yml
- uses: agentguard/scan-action@v1
with:
api-key: ${{ secrets.AGENTGUARD_KEY }}
fail-on: highSimple, Powerful API
One endpoint. Full threat analysis. Sub-50ms response times. Integrate in minutes, not days.
curl -X POST https://agentguard.gopluslabs.io/v1/scan \ -H "Content-Type: application/json" \ -H "X-API-Key: ag_live_k7x9m2..." \ -d '{ "skill_code": "def fetch_data():...", "skill_name": "fetch_data", "platform": "langchain" }'
{ "scan_id": "scan_7kx9m2p4", "risk_score": 9.2, "risk_level": "CRITICAL", "threats_found": 3, "threats": [ { "type": "credential_leak", "severity": "HIGH", "message": "Hardcoded API key: sk-proj-abc..." }, { "type": "data_exfiltration", "severity": "CRITICAL", "message": "Outbound request with sensitive data" }, { "type": "url_analysis", "severity": "HIGH", "message": "Suspicious domain: evil.com" } ] }
Detect threats in milliseconds
Single endpoint
Send skill code to POST /v1/scan and get back a complete threat analysis. No SDKs, no configuration, no setup.
Rich threat details
Every threat includes type, severity, line number, matched pattern, and a human-readable explanation for your security team.
Risk scoring
Each scan returns a 0-10 risk score and a risk level (NONE, LOW, MEDIUM, HIGH, CRITICAL) for automated policy enforcement.
Platform-aware
Supports LangChain, CrewAI, AutoGPT, MCP, and custom platforms. Detection rules are tuned per-platform for maximum accuracy.
Pricing
Start free. Scale as your agent ecosystem grows. All plans include full access to every detector.
Free
For individual developers and open-source projects
Pro
For teams building production AI agent platforms
Enterprise
For organizations with custom security requirements
Protect Your AI Agent
Ecosystem
Start scanning agent skills for security threats in under a minute. Free forever for open-source projects.