973 MCP Packages, 71% Single-Maintainer: A Practitioner’s Guide to AI Developer Security

This practitioner-focused report details large-scale risks from AI developer tooling — including high rates of vulnerabilities in AI-generated code, widespread secret leakage via MCP config files, an immature MCP package ecosystem susceptible to malicious uploads, and accelerating prompt-injection CVEs — and documents an active supply-chain compromise (UNC1069 hijacking the axios npm package and deploying the WAVESHAPER backdoor). It provides concrete mitigations (SAST on commit, secret handling recommendations, pinning/auditing MCP packages, context restrictions and network monitoring) and cites multiple studies and vendor analyses demonstrating active exploitation and systemic exposure.