---
title: "Discussion on \"Your Agent Can't Tell What It Was Told From What It Read\" | Hashnode"
description: "Discussion on \"Your Agent Can't Tell What It Was Told From What It Read\". Originally published in Securing the Agentic Stack, my weekly newsletter on the trust boundaries that appear when software can turn what it reads into action.\nYou already treat input as untrusted. You"
url: "https://hashnode.com/posts/your-agent-can-t-tell-what-it-was-told-from-what-it-read/6a8a41c871ab5aa10222609d"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [Your Agent Can't Tell What It Was Told From What It Read](https://neerazz.hashnode.dev/your-agent-can-t-tell-what-it-was-told-from-what-it-read?utm_source=hashnode&utm_medium=feed)

Originally published in Securing the Agentic Stack, my weekly newsletter on the trust boundaries that appear when software can turn what it reads into action. You already treat input as untrusted. You

Aug 23 · 8 min read · neerazz.hashnode.dev  
Read the article

## Written by

[Neeraj Kumar Singh Beshane](https://hashnode.com/@neerazz)

@neerazz · Fremont, California, USA

  
[View profile](https://hashnode.com/@neerazz)

Security Infrastructure · AI Security · Distributed Systems

Staff Security Infrastructure Engineer at Parafin, working on Zero Trust for embedded lending infrastructure with $35B+ in capital offers extended. AI security researcher: EmbedGuard, GenOps, RuntimeGuard-AI, and the PerceptFence Zenodo/GitHub support artifact. Speaker at Conf42 DevOps, Conf42 LLMs, and CIMA 2026. Gold Laureate, Hackathon Raptors Fellow, IICSPA Fellow, Sigma Xi Full Member. I write about agentic AI security, MCP/tool-runtime risk, RAG security, Zero Trust, and platform engineering. ORCID: 0009-0002-2125-1805

## Responses

Markdown supported · ⌘↵ to post

No responses yet.