Skip to content
View Danush-Aries's full-sized avatar
:dependabot:
:dependabot:
3 followers · 0 following

Achievements

Achievement: QuickdrawAchievement: YOLO

Achievements

Achievement: QuickdrawAchievement: YOLO

Highlights

Block or report Danush-Aries

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Danush-Aries/README.md

Dhanush Shankar — AI & Security Engineer

I build AI agents, then try to break them.

Portfolio LinkedIn Email

class Dhanush:
    role       = "AI & Security Engineer"
    education  = "B.Tech CSE, GITAM University (2026)"
    based_in   = "Bengaluru, India"
    builds     = ["LLM agents", "MCP tools", "RAG apps"]
    breaks     = ["prompt injection", "agent tool hijacking", "leaky system prompts"]
    certified  = ["EC-Council EHE", "Cisco Networking & Ethical Hacking"]
    looking_for = "Internships and entry-level roles in AI/ML, cybersecurity or SDE"

🛠️ Build — agents that act in the real world

Project What it does
jarvis Multi-provider AI assistant with voice, web and CLI versions; runs on local models too Python
agent-os Runtime for cooperating agents: scheduler, layered memory, tools, guardrails Python
computer-use-agent Claude controlling a sandboxed Linux desktop: screenshot → plan → act Docker
customer-support-agent Support chat that answers from a knowledge base with RAG and streams replies TypeScript
promptdiff pytest for prompts: catches when a prompt or model change silently breaks outputs CI
route-arena Typed LLM router that scores models and picks the right one per request Python

🛡️ Break & defend — security for AI and code

Project What it does
praetor Automated pentest orchestrator: recon → scan → report behind a hard scope gate, with a full audit trail CI
prompt-injection-lab Hands-on offline lab: exploit an AI agent, then defend it. No API key needed Python
mcp-sentinel Security linter for MCP configs: tool poisoning, exposed secrets, supply-chain risk CI
injectsweep Finds prompt-injection payloads hidden in documents before they reach a RAG index SARIF
tracewarden Scans recorded agent tool calls for injection, data exfiltration and RCE patterns OWASP
cve-advisor LLM-assisted CVE triage over NVD and Exploit-DB Python
More tools I've built (click to expand)

AI security: leakprobe (system-prompt leak tester) · overguard (guardrail over-refusal leaderboard) · mcp-cloak (secret redaction proxy for MCP) · mcp-audit · llm-fragility-lab · awesome-ai-agent-security

AppSec & threat intel: sast-scanner · envsentry · depdrift · yaratune · reconweave · breachintel · leakradar · network-intrusion-analyzer

Agents & dev tools: browser-automation-agent · autonomous-coding-agent · tokenwatch · agentcard · cursor-rules-forge · swara (Indic text-to-speech for blind students)

⚙️ Tech stack

Python, TypeScript, JavaScript, React, Next.js, FastAPI, Node.js, Docker, Linux, Bash, Git, GitHub Actions, PyTorch, PostgreSQL

AI: Claude API · Agent SDK · MCP · RAG · Ollama · LangChain  |  Security: Burp Suite · Nmap · Metasploit · SAST/SARIF · OWASP LLM Top 10

📊 GitHub activity

Contribution snake

Hiring for AI or security? I reply fast: danushankar3@gmail.com · or try hire in my portfolio terminal

Pinned Loading

  1. breachintel breachintel Public

    OSINT breach intelligence aggregator — forums, ransomware feeds, geo enrichment

    Python

  2. customer-support-agent customer-support-agent Public

    AI-powered customer support chat with Claude API, RAG knowledge base, mood detection, and real-time streaming

    TypeScript

  3. cve-advisor cve-advisor Public

    AI-powered CVE advisor that analyzes code and recommends patches for known vulnerabilities

    Python

  4. jarvis jarvis Public

    Portable multi-provider AI assistant — voice, web, CLI, daemon — works with zero API keys via Claude Max/Ollama; MCP-powered coding, security, and desktop automation agents

    Python 1 3

  5. mcp-sentinel mcp-sentinel Public

    Static security linter for the Model Context Protocol (MCP) trust boundary — catches tool poisoning, dangerous capabilities, exposed secrets & supply-chain risk in mcpServers configs and tool manif…

    Python

  6. praetor praetor Public

    Governed autonomous pentest orchestrator — LLM-driven recon→scan→exploit→report with a hard scope gate, full audit trail, and a blue-team detection mirror (MITRE+Sigma) for every action.

    Python