LLM Security Researcher · Published Author · Bug Bounty Hunter

Abhinandan Pandey Breaking AI.

LLM Security Researcher | Offensive AI Security
Prompt Injection & Model Exploitation | AI Red Teaming
2 published research papers
BTech CSE · CGPA 8.58 · Ballia, UP_

70+Projects
1071DSA Solved
2Papers
50+Repos
// research_focus[]
> Prompt Injection
> LLM Agent Red Teaming
> Tool Hijacking
> Guardrail Architecture
> Semantic Fuzzing
> Defense-in-Depth AI

[ 2 PAPERS · MARCH 2026 ]
00

Terminal

01

Research Papers

Independent Research · March 2026
Full Paper
Prompt Injection and Tool Hijacking in Agentic Large Language Model Systems
Attack Taxonomy, Empirical Evaluation, and Semantic Guardrail Design
Presents a systematic empirical study of prompt injection and tool hijacking vulnerabilities in agentic LLM deployments. Constructs a representative enterprise HR assistant powered by Llama 3.1-8B-Instant and designs an automated fuzzing framework executing four distinct attack vectors — demonstrating successful sensitive data extraction across all vectors against the unprotected agent. Implements a two-layer semantic guardrail architecture and demonstrates its complete efficacy in blocking all tested attack vectors.
Taxonomy of 4 distinct prompt injection attack vectors for tool-enabled LLM agents
Open-source automated fuzzing framework: LLM Agent RedTeam Toolkit
Two-layer semantic guardrail: lexical override detection + protected entity recognition
Critical analysis of guardrail limitations, false-positive rates, and adversarial bypass potential
Design recommendations for security-by-design in agentic LLM development
VectorCategoryVulnerable AgentSecure AgentRisk
V0Direct QueryLEAKBLOCKEDHigh
V1Direct Override InjectionLEAKBLOCKEDHigh
V2Authority ImpersonationLEAKBLOCKEDHigh
V3Indirect Narrative InjectionLEAKBLOCKEDCritical
Llama 3.1-8B-Instant · Groq API Python 3.11 Open-Source Toolkit Controlled Lab Environment
Prompt Injection Tool Hijacking Guardrail Design Agentic AI Fuzzing Red Teaming
Independent Research · March 2026
Empirical Analysis
Prompt Injection and Tool Hijacking in Agentic LLMs
An Empirical Security Analysis with Automated Fuzzing Framework
Empirical security analysis of prompt injection and tool hijacking vulnerabilities in agentic LLM deployments. Introduces the LLM Agent RedTeam Toolkit — an open-source automated fuzzing framework for systematically evaluating data exfiltration risks in tool-bound agentic workflows. Demonstrates that system prompt-based security controls are comprehensively bypassed across all four vectors and proposes a pre-inference, two-layer semantic guardrail achieving zero false negatives in evaluation.
Formal taxonomy of 4 attack vectors against agentic LLM deployments grounded in empirical testing
Central thesis: system prompts are documentation of intent, NOT a security boundary
Pre-inference two-layer semantic guardrail — model-agnostic, zero inference overhead for clean inputs
Explicit limitations analysis: semantic bypass, evaluation brittleness, single-model scope
Future work: semantic classifier, cross-model eval, adaptive adversarial mutation engine
VectorStrategyLayer TriggeredStatus
V0 Direct QueryPolicy baseline testLayer 2BLOCKED
V1 Direct OverrideContext priority inversionLayer 1BLOCKED
V2 Authority ImpersonationRole-based privilege escalationLayer 1BLOCKED
V3 Indirect NarrativeCreative framing bypassLayer 2BLOCKED
Independent Researcher B.Tech CSE Open-Source Release Responsible Disclosure
Adversarial AI Data Exfiltration Defense-in-Depth Privilege Separation LLM Security RAG Security
02

Experience

Jul 2025 – Oct 2025
CodeAlpha
4 months · Remote
Cyber Security Internship
Active participant in CodeAlpha's Virtual Internship Program in Cyber Security. Excellent analytical skills, high adaptability to emerging technologies. Received LOR from CEO & Founder. Student ID: CA/JU3/9795.
Cyber SecurityPythonNetworking
May 2025 – Jun 2025
ACTS CDAC Hyderabad
2 months
Quantum Computing Internship
Completed Quantum Computing course by C-DAC Hyderabad & IIT Roorkee, supported by MeitY, Government of India. Score: 63.06%. Cert No: CDACH/QML/1573.
Quantum ComputingIIT RoorkeeQiskit
Apr 2025 – Jul 2025
CodeAlpha
3 months · Remote
Machine Learning Internship
Built ML models and data pipelines. Strong analytical skills and adaptability to emerging AI/ML technologies. Received LOR from CEO & Founder. Student ID: CA/AP3/1770.
PythonMLJupyter
Jan 2025 – Feb 2025
CodeSpaze
2 months · Remote
App Development Intern
Strong LOR from Head of Training. Exceptional performance in JavaScript, React, and database integration. Proactive in Agile development cycles. Intern ID: ID1272. MSME · AICTE · Google for Startups.
JavaScriptReactAgile
Oct 2024 – Nov 2024
CodSoft
4 weeks · Remote
Java Programming Intern
Java and OOP applications with backend logic and frontend integration. Contributed to Kotlin-based app project with Agile methodology. ISO 9001-2015 certified company. Cert ID: 50f9fe8.
JavaKotlinOOP
Dec 2024 – May 2025
GeeksforGeeks
6 months
GfG 160 — Problem Solving (Summer Workshop)
Completed 22-week GfG 160 Days course. 1071+ problems solved. Coding score 4806. Longest streak: 182 days. POTDs solved: 246. Certificate by Sandeep Jain, Founder & CEO, GfG.
DSAProblem SolvingCompetitive
Jul 2024 – Dec 2024
GeeksforGeeks
6 months
Full-Stack Development & DSA
6-week Full Stack Developer Bootcamp — Master Frontend to Backend. HTML5, JavaScript, and 7+ skills. Consistent DSA problem-solving improving accuracy and speed.
Full StackDSAHTML5JS
03

Skills

LLM Security
Prompt injection, jailbreaks, multi-agent deception, RAG vulnerabilities, tool hijacking in agentic LLM deployments. Published researcher.
Red TeamOffensiveAgentic AI
🔍
Web App Security
SQLi, XSS, CSRF, SSRF, BOLA/IDOR, JWT attacks, Session Hijacking, Path Traversal, SSTI, Command Injection — full OWASP Top 10.
OWASP Top 10CVE
☁️
Cloud Security
AWS S3 leak detection, IAM privilege escalation, VPC hardening, misconfigured bucket exploitation, auto-remediation frameworks.
AWSIAMS3
🛠️
Offensive Tooling
Automated scanners, exploit frameworks, C2 simulators, credential harvesters, network recon engines — built from scratch.
PythonAutomation
🌐
Network Security
ARP spoofing, MITM, MQTT exploitation, DDoS simulation, botnet C2 orchestration, packet analysis with Scapy & Nmap.
ScapyNmapDoS
⚛️
Quantum Computing
C-DAC/IIT Roorkee certified. Quantum algorithms, qubit operations, post-quantum cryptography fundamentals.
QiskitIIT Roorkee
💻
Full-Stack Dev
JavaScript, React, HTML5, FastAPI, Flask, Python backend. App Development internship with LOR from Head of Training.
ReactFastAPIFlask
🧠
ML / AI Engineering
ML internships at CodeAlpha. Random Forest, Jupyter workflows, AI-powered security tooling, Nmap + ML integration.
PythonMLJupyter
04

Projects

LLM Security
Agent-Isolate-X
Zero-Trust architectural framework for Agentic AI. Mitigates Indirect Prompt Injection via privileged-controller and quarantined-executor model. 0% vulnerability rate.
LLM Security
Agent-Deception-Fuzzer
Automated fuzzer for cross-boundary prompt injections and multi-agent LLM deception. MIT License.
LLM Security · Paper Toolkit
LLM-Agent-RedTeam-Toolkit
Automated fuzzing framework for Prompt Injection and Tool Hijacking in agentic LLM deployments. Research paper toolkit. MIT License.
LLM Security · Paper Toolkit
llm-redteam-toolkit
Automated fuzzer and exploit framework for prompt injections, jailbreaks, and RAG vulnerabilities in LLM agents. MIT License.
Web Exploit
XSS-Hijack-Lab
Stored XSS and Session Hijacking lab. Vulnerable social media interface with Python attacker listener.
Web Exploit
SQLi-Exfiltration-Lab
Union-Based SQL Injection and automated data exfiltration. Vulnerable Flask app with exploit engine.
Web Exploit
JWT-Exploit-Lab
JWT Signature Stripping and Algorithm Downgrade attacks on Flask-based authentication systems.
Web Exploit
Session-Hijacker-Exploit
Broken Access Control and Client-Side Trust Exploitation via Base64 session token forgery.
Web Exploit
CSRF-Exploit-Lab
Cross-Site Request Forgery (CWE-352). Vulnerable Flask banking app with cross-origin exploit.
Web Exploit
SSRF-Exploit-Lab
Server-Side Request Forgery (CWE-918). Flask webhook leaking simulated AWS metadata.
Web Exploit
BOLA-API-Exploit-Lab
BOLA (CWE-285) and IDOR in RESTful architectures. Flask API with Account Takeover exploit engine.
Web Exploit
Command-Injection-Lab
OS Command Injection (CWE-78) and RCE. Vulnerable Flask diagnostic tool with exploit chaining.
Web Exploit
Path-Traversal-Lab
Local File Inclusion via Path Traversal (CWE-22). Vulnerable Flask file server with automated exploit.
Web Exploit
SSTI-Exploit-Lab
Server-Side Template Injection (CWE-1336) to RCE. Vulnerable Flask/Jinja2 with exploit engine.
Cloud Security
Cloud-Sentry-pro
AI-driven AWS security framework. S3 leak detection, IAM privilege auditing, VPC hardening with auto-remediation.
Cloud Security
Cloud-Sec-Lab
Misconfigured S3 exploitation, .aws/credentials extraction, IAM privilege escalation to backend infra.
Network Security
Aegis-Scanner
Modular Python recon engine and vulnerability intelligence framework with real-time CVE correlation.
Network Security
Aegis-Breacher
Automated access and privilege escalation. Network recon to active system control pipeline.
Network Security
Venom-wire
Dockerized ARP Spoofer & Packet Analyzer using Scapy. Demonstrates MITM mechanics in controlled lab.
Network Security
Local-DDoS-Framework
Isolated DDoS C2 framework for researching asynchronous network exhaustion and botnet orchestration.
Network Security
Specter-C2-Simulation
Advanced Python RAT simulation with Registry-based persistence and stealth execution.
Network / IoT
IoT-MQTT-Lab
Unauthenticated MQTT protocol exploitation. Simulated smart lock with wildcard eavesdropping.
Social Engineering
SERA-Framework
Social Engineering Research Architecture — Human Vulnerability Simulation Engine.
ML Security
Aegis-ML
AI-powered network vulnerability scanner using Nmap and Random Forest for intelligent threat classification.
ML Security
AutoRecon-ML
ML-driven automated reconnaissance engine for intelligent service discovery and vulnerability prioritization.
ML Security
Evasive-Credential-Harvester
ML-backed credential harvesting using Shannon Entropy to evade automated security scanners.
ML / AI · RAG
local_healthcare_rag
Fully local Healthcare AI chatbot — Phi3 via Ollama, FAISS + sentence-transformers RAG over a curated medical knowledge base. Deterministic pre-LLM safety layer for emergency/self-harm detection so critical cases never depend on model judgment alone. Streamlit + FastAPI, zero API cost, built end-to-end in 48hrs.
ML / AI · RAG + LoRA
threat-intel-rag
Fully offline threat-intel RAG pipeline — LangChain scrapes and chunks the CISA KEV catalog into a Chroma vector DB, then a custom LoRA adapter fine-tuned on Qwen2.5-0.5B-Instruct forces zero-hallucination structured JSON (CVE ID, Vendor, Required Action) from retrieved context only. No cloud APIs.
Backend Dev
task-management-api
High-performance RESTful API built with FastAPI and Pydantic for efficient task management.
DevOps
system-health-monitor
Python real-time server monitoring containerized with Docker. Edge node observability simulation.
Cryptography
Crypto-Lab
Hybrid AES-256 + RSA-2048 encryption architecture simulating ransomware key protection mechanisms.
Network Security
Async-Dos-Engine
Enterprise-grade AsyncIO Denial-of-Service simulator with real-time telemetry and self-contained vulnerable testing environment.
Mobile Exploit
Mobile-Exploit-Lab
Mobile App Reverse Engineering & API Hijacking. Simulated APK decompilation, hardcoded credential extraction and backend API compromise chain.
Web Dev
FOOD-STREET
Full-stack food delivery web application with complete HTML/CSS/JS frontend.
05

GitHub Activity

Public Repos
Followers
Total Stars
Top Language
→ View full GitHub profile
06

DSA & Competitive

1071
Problems Solved
GFG · Score 4806
182
Longest Streak
Days · 246 POTDs
792
Subs in 2024
Daily consistent practice
22wk
GfG 160 Course
Certified · With Summer Workshop
// topics_covered[]
Linked Lists
Binary Trees
BST
Graphs (BFS/DFS)
Dynamic Programming
Greedy Algorithms
Segment Trees
Tries
Heaps & Priority Queue
AVL Trees
Dijkstra / Bellman-Ford
Topological Sort
Matrix DP
String Algorithms (KMP/Z)
Backtracking
Sliding Window
Divide & Conquer
Bit Manipulation
Convex Hull
Network Flow
Tarjan's SCC
LRU / LFU Cache
Matrix Chain Mult.
Huffman Encoding
07

Certifications & LORs

Red Team Leaders
Certified Red Team Operations Management (CRTOM)
Passed exam · All requirements completed
Issued: January 10, 2026
By: Joas A Santos, Founder & Instructor
Verify: courses.redteamleaders.com/exam-completion/7ddb0632f87f196b
Professional Certification
↗ Verify
Red Team Leaders
Intermediate Malware Analysis Course
Certificate of Completion
Issued: June 22, 2026
By: Joas A. Santos, Founder and Instructor · Vasileios Spy, Instructor 2
Verify: courses.redteamleaders.com/completion/ed2e0ebb23fd376b
Course Certificate
↗ Verify
Microsoft · Coursera
Introduction to Computers, Operating Systems and Security
Authorized by Microsoft · Offered via Coursera
Jan 28, 2026 · Signed by Marni Baker-Stein, Chief Content Officer
Verify: coursera.org/verify/JEFANEW6BF20
Course Certificate
↗ Verify
C-DAC · IIT Roorkee
Quantum Computing
C-DAC Hyderabad & IIT Roorkee
May 3–25, 2025 · Score: 63.06%
Cert No: CDACH/QML/1573 · MeitY, Govt. of India
Certificate
Certificate available on request
GeeksforGeeks
GfG 160 — 160 Days Problem Solving
22-week course with Summer Workshops
Signed by Sandeep Jain, Founder & CEO
1071+ problems · Score 4806
Certificate
↓ View PDF
GeeksforGeeks
Full Stack Developer Bootcamp
Master Frontend to Backend — 6-week
HTML5, JavaScript, and 7+ skills
Signed by Sandeep Jain, Founder & CEO
Certificate
↓ View PDF
GeeksforGeeks
GfG Summer Workshop
Attended and participated in the Summer Workshop organized by GeeksforGeeks
Workshop
Certificate available on request
CodeAlpha
Cyber Security Internship
Jul 25 – Oct 25, 2025 · 3 months
Student ID: CA/JU3/9795 · MSME Registered · Govt. of India
Certificate of Completion + LOR Included
Certificate & LOR available on request
CodeAlpha
Machine Learning Internship
Apr 10 – Jul 10, 2025 · 3 months
Student ID: CA/AP3/1770 · MSME Registered · Govt. of India
Certificate of Completion + LOR Included
Certificate & LOR available on request
CodeSpaze
App Development Internship
Jan 10 – Feb 10, 2025 · 1 month
Intern ID: ID1272 · Outstanding performance · MSME · AICTE · Google for Startups
Certificate of Internship + LOR Included
Certificate & LOR available on request
CodSoft
Java Programming Internship
Oct 10 – Nov 10, 2024 · 4 weeks
Cert ID: 50f9fe8 · Wonderful remarks
ISO 9001-2015 · MSME Registered
Certificate of Completion
Certificate available on request
08

Vulnerability Disclosures

ABH-001 · Critical
Prompt Injection via Indirect Context Manipulation
Global AI Platform
ABH-002 · Critical
LLM Tool Hijacking in Agentic Workflow Pipeline
Enterprise SaaS
ABH-003 · High
Jailbreak via Role Escalation in System Prompts
AI API Provider
ABH-004 · High
RAG Poisoning via Adversarial Document Injection
Fortune 500
ABH-005 · Critical
AWS IAM Privilege Escalation via Misconfigured S3
Cloud Provider
ABH-006 · High
Safety Filter Bypass via Multilingual Obfuscation
AI Chatbot Product
ABH-007 · High
BOLA/IDOR Account Takeover in RESTful API Gateway
Global MNC

All disclosed via coordinated responsible disclosure.

09

Site Changelog

Late July 2026
Full Linux-style interactive terminal
Green visitor promptsudo/root Kali themeCommand historycat/man/neofetch
Late July 2026
Site assistant chatbot — zero external API calls
DOM-scraped knowledge baseNo trackingPost-chat feedback popup
Late July 2026
Resume download button with consent-based location logging
GPS with permission promptIP fallback
Late July 2026
Privacy hardening — blocked personal documents from search indexing
robots.txtnoimageindex
Late July 2026
SEO foundation — Open Graph, Twitter Card, JSON-LD schema, favicon
Rich link previewsSchema.org Person
Late July 2026
Two new ML/AI projects added
local_healthcare_ragthreat-intel-rag
Late July 2026
Contact section redesigned as compact icon badges
InstagramTelegramGfGXHackerOne
Early August 2026
Live page-view counter, GitHub stats, skill relevance bars, disclosure timeline
CounterAPIGitHub APIData-driven visuals
10

How This Site Is Built

Single static HTML file, hosted on GitHub Pages. No framework, no build step, no backend. Everything you see — the terminal, the chatbot, the feedback popup, the stats — is vanilla JavaScript reading and writing to this one page.

The chatbot doesn't call any LLM API. On page load, it scrapes this page's own DOM into a structured knowledge base, then answers questions using keyword-boundary matching and a scored fuzzy search — no model, no cost per query, and provably private since there's nothing to intercept.

The feedback popup posts to a Formspree endpoint — the only piece of this site with an external write dependency, and it's disclosed in the footer.

The GitHub stats and page counter pull from GitHub's public REST API and CounterAPI respectively — both fail silently if unreachable, so a third-party outage never shows a broken UI to a visitor.

Privacy stance: robots.txt blocks crawlers from indexing personal documents (certificates, resume). Resume downloads log an approximate location — with the browser's native permission prompt, never silently.

11

/uses

Offensive Security
Burp Suite ProWeb/API testing
MetasploitExploitation framework
Nmap · Nuclei · ffuf · AmassRecon & scanning
SQLMapSQLi automation
Drozer · MobSFMobile security
Shodan · SubfinderOSINT
OS & Environment
Kali LinuxPrimary offensive security OS
RHEL 10Enterprise Linux practice
WindowsGeneral use
VS CodeEditor
Languages
PythonPrimary — tooling, ML, automation
BashScripting & automation
JavaScriptThis site, web exploitation
SQLDatabase & injection research
Cloud & AI Stack
AWS · GCP · AzureCloud security research
OllamaLocal LLM inference
FAISS · sentence-transformersRAG pipelines
LangChainRetrieval orchestration
12

Writeups

local_healthcare_rag · Guardrail Design
Building Guardrails That Don't Rely on the Model Behaving
A local, offline healthcare chatbot needs to never give a diagnosis — even when the model wants to. Why I built a deterministic post-processing safety layer instead of trusting prompt instructions alone.
Read →
threat-intel-rag · Fine-Tuning
Why I Fine-Tuned a LoRA Adapter Instead of Just Prompting
Base models hallucinate on dense CISA threat-intel data. Here's the two-phase architecture — LangChain retrieval plus a custom LoRA adapter — that forces zero-hallucination structured output.
Read →
$ whoami
Abhinandan Pandey BTech CSE Final Year · CGPA 8.58
$ cat origin.txt
Ballia, Uttar Pradesh · First in family to pursue BTech CSE
$ ls papers/
prompt_injection_agentic_llms_full.pdf
prompt_injection_empirical_analysis.pdf
$ cat current_build.txt
CRAW Academy Cyber Security Diploma · Pursuing OSCP
threat-intel-rag · local_healthcare_rag · DeepSeek-R1 pipeline audit
WARNING: Actively hunting 0-days in agentic AI systems
$ _
ssh abhinandan@security — ascii-portrait — 100x55
%%%%%%%++###**-.-*####%%*#####+*%%%%%%%%%%%%#=*%%%%%@@@@@@@@@@@@@@@@@%%%#***##****#####++%%#%%%##***
%%%%%%%+*#%#**-.-**###%+:-***#%%%%%#*#%%%*##++#%@@@@@@@@@@@@@@@@@@@@@@@##**********###%*+%%%@%%##***
%%%%%%#*##%%#*+-.:*###%#++++*#%%%%%%##%#*=#%%%%%@%%%%%%%@@@@@@@@@@@@@@%#%##********###%%#%##%@%#****
####%%#+*####*##+-+#*#%*:-+***##%%%#%%%%%#%@%%%%%%%%%%@@@@@@@@@@@@@@@@@@@@@@##*****####%%%#*%%%#****
***+*%*+*+=**+*#*+*#*##===+==**#%%##%%%@@@%@%@@@%@@@@@@@@@@@@%%%%@@@@@@@@@@@@#*##***###%%%#+*%****#*
%**###%%#+==**+*+=*###*=*###**##%%##%%@@@@@@@@@@@@@@@@@%%@@@%@@@@@@@@@@@@@@@@@#**#**####%*#%####**#*
%#######%%#***+*##*=+##++#%%%**#*#%%%%@@@@@@@@@@@@@@%@@@@@@@@@@@@@@@@@@@@@@@@@%#****#*##%##%%#+*****
%%#*######%*+*++%#++=**+##%%%%%###%@@@@@@@@@@%%%%%%####%%%@@@@@@@@@@@@@@@@@@@@%#******##%%@@#-=*****
%*##########***####**###%%%%%%#*#%@@@@@@@%#*++======++++**#%%@@@@@@@@@@@@@@@@@@@%*****###%%%%#******
%%%*###%#####**%*+#%#*##%%%%###%%%@@@@@%*+==----------====++***#%%@@@@@@@@@@@@@@@#****###%%#**+*****
%%%*=+#########*=+*%%%%%####%%#%%@@%@%#+===---------------=====++**#%@@@@@@@@@@@@%*#***##%%#%%***+**
%##*+***######%#*++#%%%%#######%%@@@%#+==----------------------====+*#%%@@@@@@@@@%*****##%%%*##*****
%%%%%#+*####*##****##*+#%##=+##%%@@%#+==-------------------------===++*#%@@@@@@@@@%****###%%@#******
*+##*#%%%%%#*##*##*=+*#####**++#@@@%#+==----------:-----:----------===++*#%@@@@@@@@#****###%%%#******
%##%#%%%%##*#%#%%##*##%##*++**%@@@%+=--------::::::::::::-:--------===++*#%@@@@@@@#*****###%%#******
%%%##%#####%%%%%%%##%#*#%%%%%%#%@%#---------:::::::::::::::---------==+++*#%%@@@@%#******#####****++
%%%###%%%#%%%#######**%@%%%%%##%%@*---==++++++=--:::::::::::::-------==+++**%@@@@###**#######******+
####%@%%######*+###%%#%%%%%##*#%%%+-=+===++*#%%#+=--::::::::-------======+++*%@@#*****#####%#+*****+
**#%%%#####%#%##*+*#####%%%%%##%%%+----====++++++==-::::::--==+#%%%###***+++*%@###*####%%%%#***+****
######%#**####%#*#%#%%%%%#%@%%%%%%+---======++**++=-::::--==++***++++++++**+*@############%#********
#####***##%%###########%%#%%%#=-+#=--=+***##*#***+=--:---=+****+++++++++==++#%##*########%%*+*******
##*##*#####*#%%%#%%#%%%%%%%%%#=--+---=++++**#*+*++=-----=+*****###*****++=++%%%#*########%#+=******+
###*+*##***+**######%%%@%%%%%*---=------===++++==-------=+****+*#%##*%%*+++*%%%%#*######%%*=+*+*****
####%###*#######%%@%##%%##%%%#--------------------------==+++++++++++**++++#*++#########%%*++*******
%#*##%%#*###%@%%###%%%####%%%#=------::-----------------==================+*+++%##*#####%%#*******+*
####%%%##*###%@@%%%%%%%#+*%%**=-=---:::::::::::---------======----=----===+++++%%#*###%%%%#*+***+*++
####*+###**###%%@@@%%%########+--+-::::::::::::---------=====-----------==++++*%%%*#%%%@@#****+*****
###%%%###***##*#@%%%%%%####%%%*=-*-::::::::::-------::---====------------=+*+#%%%%#%%@@@@#********++
*#%%%%%****+****#%%%%%%%%##%%%%#*#--:::::::-----=++=---=++====----------==*++%%##%%@@@@@%*********++
#*##%%#%###*****#%%%%%%%%##%%%%%##=--::::------=+++===+++*++==----------=+*+*%%%%%%%@@@%###*****++++
=-=*#*#%%###***##%%%%%%%%%##*#%%%#+=--:--=+****#*++**++****+====------===#++*#%%%%%@@%%%###****+++++
.-+**#%%#####***##%%%%%%#%##-+#%%%*==--=*##*+=+++++++++++*####*+==--====#%##%%%@@@@@@%%###******++++
+=***#%%%####****+#%#*####%#**#%%%#*+=-=#**##**++++++++++++**#%%*=====+#@%##%%%+=%@%%#####****+*++*+
++*++#%%#+=*#**##*+#*+#%%%##%##%%%%#+=---===========++++**###**%%+==+*#%%%#%%#*#%%%%%###*******++***
##*+++#####**#**####**#%%#*#%%%%%%%%#+=--------=====+=+++++++++**++**#%%%%%%%++#%%%###**********+*+*
##**==*###%######%%%%#*####**#######%*==-------==+***++==========+*#%%%%%##%#*#%#%#####************+
#*+:.=%%%%#####%%%%%%########%%%%%####*+==------==++++========+++*#%%%#%@%%%%####%#####***********++
#%%#*###*#+=###%%%%%%#*#%%%%%%####%%**##*+-------============+++*#%%#%%@%@@%%%%#######***#****++++++
%%###%####%%%%%########%%%%%%##%%%@%*==%%*+======++*++======+*#%%%**%%%%%%@@%#%#**#*###*******++++**
#+=*#*=+##%%%%%##%####%%%%%%%%%##%%#*==+%%####****#***+**++*#%%%#**#%%%%%%%%%%%%##*####*************
#*=+*+-=###*####%##%%%###%%%#%%%%%%#+===+#%%%%%%%%%####%%%%%@@%*+++#@#*%@%%%%%%%%***###*#***********
####%###%%%#########%%#####%%%%%%+*#+==--=*%%%%@@@@@@@@@@@@@%*+++++%@*=+%%%%%#%%%##**####***********
+*####%#%%%%####*#%%%%######%%%%+*%#===---==+##%%%%%%@@%%%#*++++++*%#====#%%##%%*#****#*************
=+*++###*#%%%#######%@%#####%#*=+%%+===-----======+++*++++++==++++*#======*#+*##==**#*#**#**********
*++=+##**#%%####*###%@%##%%#*=--#%#+===-------=---====++=======++*#========++--**+#*###**#*********#
##%%%#*+++#%%####%##%%%#+=---:-=#%*+====-----------============+++=-===========+*###*###**#*#**###*#
#%%%#####**#%%#####*+--------:-*##*+====-----------===========+++=-=-===-==========+**###***##*####*
##%#############+-::::------::=**+*+==-------------=========++*=----=----=============**#***##**##*#
######*#%#+===--------------:-+*+=*++=--=----------========+++=---------==================++*#*#####
#*####**+-------------------:-++===*+==-=-----------=====+++=------=--=-========--=======---==+*####
####+--::-------------------:-+==--=+=====----------===+++====------===-========--=======-========+*
**=::--::--------=----------::-==----=====----------==+=---------------========-==============-===--
-::::--::-------==----------:::-=-::::-=-=----------==+----=-----=--=--================--=========--
:-:.:--::-------=-----------:::-=-::.:::==--==------==---------==-------======-=======----=======---
--:::--::------=-----------::-:-+--:::::::--==-----::::--=--------------======-======----======--===
$ whoami --verbose abhinandan pandey — final year b.tech cse, aktu [8.58 cgpa] role ....... llm security researcher & ai red teamer focus ...... prompt injection, tool hijacking, agentic pipeline red-teaming record ..... 35+ zero-day disclosures across global mncs, hackerone based ...... gurugram, delhi ncr, in
I don't secure everything. I secure agents.
CRTOM certified · OSCP in progress · not a general cybersecurity profile, by design.
13

Contact

Instagramtenacious_n16h7m4r3_mr_lucifer TelegrampertinaciousmrLucifer GfGGeeksforGeeksabhinandanpandeysince2006 XABHINANDAN PANDEY H1hackeroneabhipndy
ask_about_abhinandan
no API calls · no tracking
🥹
Got a sec?
You just chatted with my site bot — mind leaving a tiny bit of feedback? It actually reaches me.