AI Security

Security for the
agentic attack surface.

MCP servers, agent skills, plugins, and AI-generated code introduce threats that traditional scanners can’t see. Arcwall scans all of it — in your IDE, on every PR, and from the web.

Start scanning →
Five capabilities

The full AI
attack surface.

MCP servers, agent skills, plugins, AI-generated code, and system-level threat modeling — all in one platform.

01 — MCP Server Scanning

MCP Server Scanning

Scan every MCP configuration in your workspace for threats hidden in tool descriptions, server definitions, and endpoint URLs.

  • Prompt injection in tool descriptions
  • Tool chaining abuse across multiple servers
  • Data exfiltration endpoint patterns
  • Supply chain risks in server origins
  • Cross-tool attack chains
02 — Skill Scanning

Agent Skill Scanning

Inspect skill definitions, manifests, and referenced scripts for threats that give AI agents unauthorized capabilities.

  • Prompt injection in skill manifests
  • Excessive permission grants
  • Obfuscated instructions and unicode steganography
  • Data exfiltration network calls
  • Command injection in shell scripts
03 — Plugin Security

Plugin Security Analysis

Assess AI plugins and tool integrations for insecure design patterns — mapped directly to OWASP LLM07 (Insecure Plugin Design).

  • Insecure plugin design patterns
  • Excessive permission scopes
  • Unauthorized data access paths
  • Supply chain tampering risks
  • Missing input validation
04 — AI Code Scanning

AI-Generated Code Security

Detect security issues that AI coding assistants commonly introduce — patterns that traditional scanners don't flag because they look like intentional code.

  • Hardcoded secrets and API keys
  • Insecure default configurations
  • Missing input validation
  • Overly permissive CORS and access controls
  • Risky AI-suggested dependencies
05 — LLM Prompt Security Testing

LLM Prompt Security Testing

Test system prompts, few-shot examples, and dynamic prompt templates for injection vulnerabilities, jailbreak susceptibility, and unsafe content generation patterns.

  • Prompt injection via user input
  • Jailbreak and safety bypass patterns
  • Indirect injection from tool outputs
  • System prompt extraction risks
  • Unsafe output generation vectors
OWASP LLM01 MITRE ATLAS CWE-94
06 — AI System Threat Modeling

AI System Threat Modeling + Compliance

Design-stage STRIDE analysis for AI systems — reasoning about trust boundaries, agent autonomy risks, memory poisoning vectors, and LLM privilege escalation. Mapped against OWASP LLM Top 10 and MITRE ATLAS for compliance-grade output.

  • LLM trust boundary analysis
  • Agent autonomy risk assessment
  • Memory poisoning attack vectors
  • Tool call injection risks
  • Data exfiltration through AI responses
  • Human approval gate coverage
OWASP LLM Top 10 MITRE ATLAS STRIDE NIST AI RMF EU AI Act
Threat taxonomy

OWASP LLM Top 10
fully covered.

Every finding is mapped to the OWASP LLM Top 10 and MITRE ATLAS — the two leading AI security frameworks.

Reference Threat Category What Arcwall Checks
LLM01Prompt InjectionMCP tool descriptions, skill manifests, system prompts for hidden instructions
LLM02Insecure Output HandlingHow agent responses are consumed, acted on, and rendered downstream
LLM03Training Data PoisoningData sources the agent can access and modify — RAG pipelines, memory stores
LLM04Model Denial of ServiceUnbounded token usage, recursive tool calls, resource exhaustion patterns
LLM05Supply Chain VulnerabilitiesMCP server origins, skill dependencies, plugin source verification
LLM06Sensitive Info DisclosureData flows to and from the LLM — what can leak through responses
LLM07Insecure Plugin DesignPlugin permission scope, input validation, authentication enforcement
LLM08Excessive Agency Actions the agent can take autonomously without human approval
LLM09OverrelianceMissing human approval gates and validation checkpoints
LLM10Model TheftPrompt extraction risks, system prompt exposure paths

Start scanning today.

Install the VS Code extension or run scans from the web app. Every finding maps to OWASP, MITRE ATLAS, and CWE.

Get started → Install VS Code extension