By Vero SecurityBuilt and funded by Vero Security, the AI Security Engineering platform.Read the charter

Coding agent exploits

Coding agents read repositories, issues, and dependencies while holding shell access and developer credentials, so a hidden instruction can become code execution. Study the public exploit chains against Copilot, Claude Code, Cursor, Devin, Kiro, and others to see how small permission gaps turn into full compromise.

Read in this order

  1. ArticleGitHub Copilot: Remote Code Execution via Prompt Injection (CVE-2025-53773)Start here: a full chain from injected text to code execution, with a CVE.
  2. ArticleTrail of Bits: Prompt injection engineering for attackers: Exploiting GitHub CopilotHow attackers craft payloads that survive real agent behavior, from a security firm.
  3. ArticleAgentic ProbLLMs: The Month of AI Bugs 2025A month of disclosed bugs across many coding agents, to spot recurring patterns.

Save your progress

We email you a sign-in link. No password.

By signing in you agree to the terms. The privacy policy explains how we use your details.

Join the AI security engineering community

Free. Roadmap updates, chapters near you, and first word when credentials open.

We’ll email you community updates. Unsubscribe anytime. By sending this you agree to the terms. The privacy policy explains how we use your details.