Vercel's internal security agent audits every line of code; rauchg says don't rely on prompts
alvelda · x · 2026-11-05
Vercel CEO Guillermo Rauch unveiled that the company's internal security agent (nicknamed @hacker) automatically audits every single line of code, and the capability is now available as a product that works even for repos not hosted on Vercel. His core argument: you can't count on engineers having a skill installed or prompting "please make it secure" — put the check in the pipeline, not in the prompt. The system was built entirely by Vercel's security team.
The post also points to Agent Field Kit, a paid collection of 20 security templates for teams running AI agents ($29 each, $79 bundles), covering secrets rotation, supply-chain defense, and inference key abuse caps, with free incident steps for the Oct 21 Railway env-var exposure and LiteLLM security incident.
More from coding & agent
- A three-step Codex + Claude Code workflow for building websites with no obvious 'AI slop' — randomname11111_ · 2026-11-05
- Self-Harness: Shanghai AI Lab's Method Lets Agents Fix Themselves, 30.2%→84.6% Without Retraining — Promptmethus · 2026-11-05
- Jeff Dean's Discovery Loop Ran 400 Interviews in 3 Weeks, With ~20 Automation Loops — salgar · 2026-11-05
- Users report Claude Code subscription quota drains much faster when accessed via T3 Code — yihui_indie · 2026-11-05
- TokenFuel Launches Sponsor Pools to Reward Developers With AI Tokens — RichardsonDx · 2026-11-05
- gstack /review catches 22 of 84 blocking bugs vs 3 for plain Opus, evals show — garrytan · 2026-11-05