Posts tagged “AI Agents”
Your auth check passed. Now prove what the agent did.
An agent calls your API with a valid token, the scopes allow it, and the write lands with a clean 201. Nobody asked for it. Here's how to fix authorization for agents first, then build an audit trail that shows exactly what the agent did.
Catch secrets before they reach git: a pre-commit check for Node repos
GitGuardian found commits co-authored by Claude Code leaked secrets at 3.2%, twice the 1.5% baseline. A 10-minute pre-commit secret scan for Node repos, a CI backstop for --no-verify, and what to do when a key slips through anyway.
Stop user enumeration on your Node.js login, signup and reset endpoints
AI agents made hundreds of millions of requests to login pages in the first half of 2026. If your API answers differently for real and unknown emails, it hands them a target list. Here's how to make login, reset and signup give one answer.
Lock down npm installs when an AI agent adds your dependencies
In late October, 96 malicious npm packages appeared in three days, several running code at install time. How to turn off install scripts in npm and pnpm, allow the few builds you need, and review every dependency an AI agent adds.
Your MCP server is an API. Debug it like one.
A remote MCP server works on localhost, then fails in production with a vague "tool call failed." Lost sessions behind a load balancer, expiring OAuth tokens and proxies that cut streams off are all visible at the HTTP layer. Here's where to look.