TLDR Tech / InfoSec
Agent Skills are the New Packages of AI: It's Time to Manage Them Securely
JFrog launched the Agent Skills Registry, a centralized repository for governing AI agent skills — reusable, file-system-based instruction sets that execute with the invoking user's privileges and carry the same supply chain risks as OSS packages: prompt poisoning, malicious code, version drift, and weak provenance. The registry integrates with Agent Skills, ClawHub, and OpenShell, and enforces a publish-time security pipeline via `jf skills publish` that runs a two-stage behavioral scan, generates in-toto compliant attestations, and optionally produces cryptographically signed provenance evidence before a skill reaches the repository. At install time, `jf skill install` verifies that evidence chain, providing zero-trust consumption across coding agents, CI/CD pipelines, and automation tooling without locking teams into vendor-specific marketplaces.
Read the full story