AI agent security checklist for developers
A practical, reviewable checklist for the controls that matter before an AI agent can touch real code, systems, or customer data.
Read guideBlog
Security notes for AI-assisted development.
Practical guides on prompt injection, MCP risk, agent permissions, approvals, and action logs.
Featured
Latest guides
How to detect prompt injection in AI-generated code
Trace untrusted instructions from their source to the model and from model output to every execution-capable sink.
MCP security risks for Claude Code users
An MCP server extends what Claude Code can reach. Review identity, scopes, commands, network access, and approval behavior before connecting it.
Vibe Coding Security: The Complete Guide to Securing AI-Generated Code
A practical guide to the code, permission, tool-use, and MCP checks that keep fast AI-assisted development reviewable.
Pre-Write Scanning vs SAST: Why Static Analysis Misses What AI Agents Actually Do
SAST analyzes static artifacts. Pre-write checks add a decision point before an AI agent writes risky code or executes a sensitive action.
AI Agent Permission Policies That Actually Work
Replace “the agent has tools” with explicit allow, block, and approval rules that check identity, action, target, environment, and impact before execution.
Secret Leaks in AI-Generated Code: How Agents Expose API Keys
AI coding agents can hardcode keys, read .env files, paste tokens into chat, and commit credentials. Catch secret leaks before code lands and before agents open secret files.