---
title: "Vishal Solanki (@vishalsolanki) | Hashnode"
description: "Full-Stack Developer · Next.js, React, Node.js · Noida, India. Read the latest articles by Vishal Solanki on Hashnode."
url: "https://hashnode.com/@vishalsolanki"
---

@vishalsolanki

# Vishal Solanki

@vishalsolanki · Noida, India · Joined October 2026

Full-Stack Developer · Next.js, React, Node.js · Noida, India

[Website](https://www.vishalsolanki.online/)

## Comments

[Thanks Abraam, both additions are right. I checked the Codex docs and you're correct about AGENTS.override.md: in each directory from the project root down to the working directory, Codex looks for it before AGENTS.md. I'm adding it to the find command and the CODEOWNERS list. Running the check in CI on every PR beats my one-time read, since submodules and vendored folders arrive later. On your question, I only have the CSO Online write-up to go on. It describes tests with the file in Downloads and says the folder also had to be trusted for Codex. It doesn't describe a test with a file nested inside a trusted repo. An analyst quoted there says the technique could be adapted to cloned repos, shared drives and package contents, but that's commentary, not a test result.](https://hashnode.com/posts/ai-coding-agent-prompt-injection-claude-md-agents-md/6ae0f55d40406a1a227f9be4/comment/6ae86e9367ff2256c81c3dad#reply-6af9d0cf205d4e295f8b34cf)

Reply · Article · 2d ago · [Prompt injection in AI coding agents: how one CLAUDE.md or AGENTS.md can hijack them](https://hashnode.com/posts/ai-coding-agent-prompt-injection-claude-md-agents-md/6ae0f55d40406a1a227f9be4)

[Fair point, Adam. A stale file is the more common failure, and it looks like a config bug because it is one. CODEOWNERS is defence 3 in the post for that reason: a hostile edit and a forgotten update go through the same review. One habit helps with staleness. When a PR changes the test or build command, the reviewer checks AGENTS.md and CLAUDE.md in that same PR.](https://hashnode.com/posts/ai-coding-agent-prompt-injection-claude-md-agents-md/6ae0f55d40406a1a227f9be4/comment/6aee4b63e2824a7aaea278ff#reply-6af9d0ae1f9628f006a7694c)

Reply · Article · 2d ago · [Prompt injection in AI coding agents: how one CLAUDE.md or AGENTS.md can hijack them](https://hashnode.com/posts/ai-coding-agent-prompt-injection-claude-md-agents-md/6ae0f55d40406a1a227f9be4)

[Thanks Charlie, that's a better format than my plain log. One row per question and one column per assistant makes the month-to-month changes easy to see. The cited sources column is the useful part: if the same 2 or 3 directories keep showing up, that's where to get listed next. Agreed on saving the exact wording too. Otherwise you can't tell if the answer changed or the question did.](https://hashnode.com/posts/get-cited-by-chatgpt-perplexity-ai-search/6ae0f56934b2ef554aff7de1/comment/6ae3fa56ca0d0190fc2a6598#reply-6af9d07d205d4e295f8b3310)

Reply · Article · 2d ago · [How to get your website cited by ChatGPT, Perplexity and Google AI Mode](https://hashnode.com/posts/get-cited-by-chatgpt-perplexity-ai-search/6ae0f56934b2ef554aff7de1)

[Hi Tarun! You're right - trusting the first X-Forwarded-For value without verifying how the proxy handles it can make the rate limit spoofable. The correct approach is to use the hosting platform's trusted client-IP header or derive the IP from a properly configured trusted proxy chain. Pairing IP-based limits with account or email-based limits for OTP routes adds another layer of protection. I'll clarify this in the example.](https://hashnode.com/posts/secure-nextjs-server-actions/6ae0f5a1e8645c6ca9f59a08/comment/6ae1f0e6ec3b6bb0091bc8ca#reply-6ae2112bb78e490adba0e5b2)

Reply · Article · Oct 28 · [Securing Next.js Server Actions: a 10-point checklist for 2026](https://hashnode.com/posts/secure-nextjs-server-actions/6ae0f5a1e8645c6ca9f59a08)

[Hi Damiano Good observation! AWS reported a PHP2Shell exploitation attempt as early as October 17, four days before public disclosure. So at least some exploitation attempts were already underway before the leak. What we can't establish yet is whether the leak accelerated subsequent attacks or expanded the number of attackers involved. Cloudflare did observe attackers modifying payloads to bypass its protections, but that alone doesn't prove the leak caused the change. It highlights an important challenge with coordinated disclosure: even when the process is carefully planned, defenders can't assume the vulnerability will remain secret until patches are available.](https://hashnode.com/posts/php2shell-cve-2026-109618-fix/6ae0f595e8645c6ca9f57be6/comment/6ae2020e46d69ef92fd3e318#reply-6ae20f6f80bde456d92163de)

Reply · Article · Oct 28 · [PHP2Shell (CVE-2026-109618 and CVE-2026-109619): is your website affected, and how to fix it](https://hashnode.com/posts/php2shell-cve-2026-109618-fix/6ae0f595e8645c6ca9f57be6)
