AI Security

Practical security for AI agents and prompts

AI agents get real tool access, real credentials, and real blast radius. This is practical prompt security and agent safety for teams shipping AI features — reviewing prompts, tool access, and risky instructions, and catching trust-boundary problems in CI before they reach production. No hype, no fear.

Get the prompt security checklist → Subscribe for updates ↓

Built by Myles Agnew · Prompt security · Agent safety · CI checks · Sydney, AU

Start here

AI Agent Prompt Security Checklist →Review AI-agent prompts, tool access, risky instructions, and CI-friendly prompt checks — a practical starting point. promptgenie →Prompt linting, static risk scanning, testing, and CI workflows for safer AI-agent prompt pipelines. promptgenie on GitHub ↗The open-source project — integrate it into CI to catch risky agent prompts before they ship.

What to review

Prompt injection and untrusted-content trust boundaries
Tool access — least privilege for what an agent can call
Risky instructions and unsafe default behaviours
Secrets and credentials reachable from a prompt
Output handling and downstream action safety
CI-friendly prompt checks that run on every change

Get prompt-security templates and updates →

Related focus areas

Security →Website, DNS, cloud, and WordPress checklists and baselines. Automation →CI checks and workflows that run prompt security on every change. Security Tools →Open-source tools and calculators you can run yourself.

Follow the Tools

Ship AI features without shipping AI risk

Get the prompt security checklist, promptgenie updates, and practical AI-agent security patterns as I build them.

Join the list → Get the checklist ↓