A 16-year-old built AegisKernel to block hijacked AI agents before they run

Adi Kshatriyag, a 16-year-old solo builder, released AegisKernel, a tool that enforces deterministic capability checks on LangChain, CrewAI and AutoGen agents before a tool executes, rather than relying on a second language model as a probabilistic guardrail. The project, surfaced on Hacker News, targets builders who grant agents shell, file and database access and want explicit permission grants verified before each tool call.

A 16-year-old built AegisKernel to block hijacked AI agents before they run Kshatriyag, a 16-year-old solo builder, is selling deterministic tool checks for LangChain, CrewAI and AutoGen agents. By Ryan Merket https://runtimewire.com/author/ryan-merket · Published Primary source: Hacker News https://adikshatriyag-crypto.github.io/aegis-kernel/ Why it matters Agent builders are giving models shell, file and database access while still relying on probabilistic guardrails to catch abuse. AegisKernel tests a stricter idea: explicit capability grants checked before a tool runs. Adi Kshatriyag is asking a question that gets sharper every time a coding agent receives terminal access: why should a second language model get to vote on whether the first one can run a dangerous command?… When you purchase through links in our articles, we may earn a small commission https://runtimewire.com/affiliate-monetization-standards . This doesn’t affect our editorial independence.