Research Paper: Systematic Static Analysis of AI Agent Frameworks

A 39KB academic paper documenting our scan of 10 major AI agent frameworks.

Highlights

• 11,036 files scanned, 6,173 findings, 1,764 critical
• 22 detection rules with CWE/CVSS mappings
• 6 novel attack vectors beyond OWASP ASI Top 10
• Comparative analysis: Bandit & Semgrep found 0 agent-specific findings vs AgentGuard's 6,173
• 21 academic citations

Sections

1. Abstract, Introduction, Methodology
2. Results with per-framework breakdown
3. 6 Key Findings (ASI-AGENT-COLLUSION, CWE-1188, Memory Poison, etc.)
4. Discussion, Conclusion, References
5. Appendices A-B

Ready for Hacker News / Reddit / academic preprint.