Skip to content

Agent Audit v0.15.1 Security scanner for AI agents and MCP configurations based on OWASP Agentic Top 10 (2026)

Choose a tag to compare

View all tags
@HeadyZhang HeadyZhang released this 06 Feb 11:06
· 48 commits to master since this release
v0.15.1
4458c62

What It Is

CLI static analysis tool — "ESLint for AI agents". Detects vulnerabilities in LangChain, CrewAI, AutoGen, and MCP server
configurations before they reach production.

Key Stats
┌────────────────┬─────────────────────────────────┐
│ Metric │ Value │
├────────────────┼─────────────────────────────────┤
│ Rules │ 40 built-in rules │
├────────────────┼─────────────────────────────────┤
│ OWASP Coverage │ 10/10 ASI categories │
├────────────────┼─────────────────────────────────┤
│ Output Formats │ Terminal, JSON, SARIF, Markdown │
├────────────────┼─────────────────────────────────┤
│ Python Support │ 3.9, 3.10, 3.11, 3.12 │
└────────────────┴─────────────────────────────────┘
What It Detects

  • ASI-01 Goal Hijacking: Prompt injection, missing goal validation
  • ASI-02 Tool Misuse: Command injection, SQL injection, unsanitized inputs
  • ASI-03 Privilege Abuse: Excessive permissions, long-lived credentials
  • ASI-04 Supply Chain: Unverified MCP servers, unpinned dependencies
  • ASI-05 Code Execution: Unsandboxed eval/exec, data exfiltration chains
  • ASI-06 Memory Poisoning: Unsanitized memory writes, unbounded history
  • ASI-07 Inter-Agent Comms: Unencrypted channels
  • ASI-08 Cascading Failures: Missing circuit breakers, no error handling
  • ASI-09 Trust Exploitation: Missing human approval, impersonation
  • ASI-10 Rogue Agents: No kill switch, self-modification risk

Quick Start

pip install agent-audit
agent-audit scan .
agent-audit scan . --format sarif --output results.sarif --fail-on high

Links

Assets 6
Loading