I build practical tools for autonomous-agent security, prompt-injection defense, and research automation.
My work is focused on making AI agents safer, more inspectable, and more useful in real workflows — especially where they touch tools, data, credentials, code, markets, or external systems.
-
Autonomous-agent security
Controls, checklists, and test cases for safer agent deployments. -
Prompt-injection defense
Practical ways to detect, benchmark, and reduce tool misuse, data exfiltration, approval bypasses, and persistence risks. -
Research automation
Local-first workflows for diligence, signal discovery, investing research, and structured decision support. -
Agent tooling
Small, composable utilities that help agents lint configs, audit behavior, run repeatable checks, and produce better artifacts.
-
agent-security
Security-focused agent skillpack and audit helpers. -
agent-config-linter
Dependency-light linter for unsafe autonomous-agent configuration patterns. -
agent-security-bench
Benchmark cases for prompt injection, tool misuse, data exfiltration, persistence, and approval-bypass scenarios. -
ticker-due-diligence-cli
CLI for structured, leading-indicator-focused stock diligence notes.
ai-security · agent-security · prompt-injection · autonomous-agents · research-automation · investing-tools · local-first-ai
I like tools that are:
- small enough to understand
- easy to run locally
- explicit about failure modes
- useful before they are “platforms”
- designed around real operator workflows
Best way to reach me is through GitHub.