Security-Aware Specification for Agent Skills

Design a security-aware Agent Skills specification that preserves the framework’s flexibility while providing meaningful security properties across compliant implementations.

Background

The current Agent Skills specification imposes no mandatory security properties, leaving gaps around capability declarations, dependency pinning, and content integrity. Overly prescriptive requirements could harm expressiveness, while permissive ones fail to reduce risk.

The authors call for a specification-level redesign that balances flexibility with enforceable security guarantees through coordinated ecosystem engagement.

References

The challenge of designing a security-aware Agent Skills specification that preserves the framework's flexibility while providing meaningful security properties is an open problem that requires engagement from the security research community, the agent platform developers, and the Skill author community.

— Towards Secure Agent Skills: Architecture, Threat Taxonomy, and Security Analysis  (2604.02837 - Li et al., 3 Apr 2026) in Section 7.2, Open Challenges (C7: Specification-Level Security Properties)

Explicitly considering and representing security features is an open challenge for researchers and tool builders, especially while agentic practices are still forming and unstructured.

— The Security Feature Location Problem  (2609.04899 - Hermann et al., 4 Sep 2026) in Section 1, Introduction

Which implementation characteristics indicate the presence of a security feature, and how do these characteristics relate to the security properties it helps enforce?

— The Security Feature Location Problem  (2609.04899 - Hermann et al., 4 Sep 2026) in Section 5, Research Challenges, subsection “C1.2: What are the characteristics of security features?”