Chainguard Launches Hardened AI Agent Skills to Secure Software Development

Event summary

  • Chainguard introduced Agent Skills on March 17, 2026, a continuously maintained catalog of hardened AI agent skills.
  • The platform automatically reviews, hardens, and publishes skills with full audit history to prevent supply chain attacks.
  • Recent attacks involved 39 malicious skills and 2,200 variants targeting OpenClaw registries, delivering the Atomic macOS Stealer (AMOS).
  • Agent Skills will expand to include proprietary skills and custom policy configurations later in 2026.

The big picture

Chainguard's move addresses the rapidly expanding attack surface in AI-driven software development. As AI agents become integral to development workflows, the proliferation of unvetted skills poses significant supply chain risks. This launch positions Chainguard to capitalize on the growing need for secure AI artifacts, aligning with broader industry trends toward securing the software supply chain.

What we're watching

Adoption Pace
How quickly developers and enterprises will integrate Chainguard Agent Skills into their AI-driven workflows.
Security Effectiveness
Whether the hardening process can effectively prevent emerging supply chain attack vectors.
Market Expansion
The pace at which Chainguard can expand its catalog to include proprietary skills and broader repository coverage.

Related topics

Technology