Building Autonomous Agent Firewalls, Adversarial Security Infrastructure & High-Velocity Dev Tools
╔══════════════════════════════════════════════════════════════════════════════╗
║ "Autonomous AI agents need deterministic security firewalls, not thoughts ║
║ and prayers. If an agent can execute code, it must be boundary-checked." ║
╚══════════════════════════════════════════════════════════════════════════════╝
The In-line Zero-Latency MCP Security Firewall & 100-Probe Red-Teaming Suite for Autonomous AI Agents.
Intercepts tool execution payloads, terminates indirect prompt injection before execution, enforces strict SSRF egress policies, and isolates agent runtime privileges.
- Companion Defense: Tchapwo1/tcha-claude-skills — Deterministic transport firewall for Claude Code & Desktop MCP tools.
- Adversarial Scanner: Tchapwo1/agent-redteam — 7-vector automated vulnerability scanner for agent system prompts.
Viral, zero-friction engineering tools designed for the modern AI and Cursor ecosystem:
|
The Superpower Skill Suite for AI Coding Agents. Drop-in engineering skills for Cursor, Claude Code, and Windsurf. Slashes code bloat by 60%, draws ASCII architecture diagrams, and enforces Apple craft. |
🕵️ RepoSpyForensic SaaS Reverse-Engineering Engine. Sniffs any web application in under 1 second, extracts underlying frameworks, design tokens, and generates instant Cursor prompts to clone the architecture. |
|
Autonomous AI Agent Vulnerability Scanner. Audits system prompts and agent pipelines against indirect prompt injection, tool privilege escalation, memory poisoning, and SSRF egress. |
🔥 GitRoastSavage Forensic Code & Hygiene Roaster. Analyzes any public GitHub repo's commit debt, stale PRs, and dirty code to compute a Survival Score with 1-click Twitter/X shame cards. |
Core Architecture: [Zero-Trust Agent Security, In-line MCP Firewalls, Adversarial Red-Teaming]
Languages: [TypeScript, JavaScript (ESM), Rust, Python, SQL]
Infrastructure: [Cloudflare Workers & D1, Docker Containers, Edge Telemetry, GitHub Actions]
AI Engineering: [Claude 3.7 Sonnet, Cursor Rules, Multi-Agent Swarms, Prompt Hardening]