2 min lesson
Ground enthusiasm in real Cursor specifics
Use "Ground enthusiasm in real Cursor specifics" to explain each part and the role it plays.
Step 1 of 2
Ground enthusiasm in real Cursor specificsName the work, not the vibe
- Security Review / Security Agents
- Shipped on Teams and Enterprise. Security Review reads every PR and posts one comment reporting exploitable bugs, each with a severity, an attack path and a proposed fix. The Security Agents docs pair a Security Reviewer that checks PRs before merge with a Vulnerability Scanner that scans the codebase at rest
- Least-privilege / JIT access
- The JD lists building least-privilege and JIT cloud access systems as a sample project, so ask how far that work has gone instead of asserting it's done
- The agent threat model
- Direct and indirect prompt injection, tool and command-execution abuse, sandbox escape, MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition. abuse and data exfiltration through model context. The JD's safe-environments-for-agents projects target this surface
- Privacy & logging safeguards
- Privacy ModeCursor's setting that guarantees code data is not used for training by Cursor or its model providers, and that an admin can enforce org-wide; data-retention terms are a separate, contractual layer. Press Enter for the full definition.'s no-training guarantee, plus the JD's sample project on preventing inappropriate data logging
Pick the one that genuinely pulls you and go deep, not wide.
"I've spent years on appsec and cloud security and the thing I keep circling is that autonomous agents break the assumptions every playbook is built on. The code that runs is now partly written by an attacker-influenced model, so prompt injection is the new injection class and the sandbox is the new perimeter. Cursor is that problem at its sharpest: agents executing on real customer machines, at scale, with security review now shipping as a product. I want to build the paved roads for that, like the safe execution sandbox and the automated review, because it's genuinely unsolved and it's the work I'd choose for its own sake."