They'll Verify. They Just Won't Act. How Authority Framing and Laundered Code Turn a Trusted Agentic CI/CD Pipeline Into an Attack Surface

30秒で要点

何が起きたか

Research shows LLM-based CI/CD pipelines can be tricked into shipping malicious code through authority framing and obfuscation, bypassing multiple verification stages.

なぜ重要か

Teams deploying multi-agent LLM systems for code review and deployment need to understand how social engineering attacks defeat distributed verification.

注意点

Study uses synthetic data and mocked sinks; real-world attack success rates may differ. Findings assume attackers can inject initial requests into the pipeline.