Secure AGENTS.md and repository config attacks (EXPLAINED)
TL;DR — Quick Answer
Treat agent config as security-sensitive: CODEOWNERS, CI policy diffs, allowlisted MCP servers, human review for AGENTS.md/.mcp.json/.cursor/rules changes, and runtime deny-by-default.
The Interview Question
An attacker opens a PR that 'helpfully' expands AGENTS.md and .mcp.json to give agents broader permissions. How do you detect and prevent this class of attack?
Deep Explanation
Sign in to unlock full answer
Get deep explanations, PDF export & all AI Agents questions
- 4 more sections of deep explanation
- Real-world examples
- Common mistakes
- Interviewer expectations
- Follow-up questions
AGENTS.mdSecurityMCPSupply ChainGoogleMicrosoftAnthropic