CoSA: The Conflict of Skills Attack in Real-World LLM Agent Ecosystems

# CoSA: The Conflict of Skills Attack in Real-World LLM Agent Ecosystems

https://gcdradar.vercel.app/incidents/cosa-the-conflict-of-skills-attack-in-real-world-llm-agent-e · AIR-2026-1568 · closed 2026-11-10

- Lane: Research
- Severity: 1 (Research / proof of concept): Demonstrated by researchers with no known real-world exploitation.
- Fix status: Fix status unknown
- First reported: 2026-10-27
- AI role: Attack on AI
- Products: LLM agent ecosystems

## Summary

Researchers identify the Conflict of Skills attack mechanism where LLM agent skills reuse artifacts without preserving necessary authorizations. The study demonstrates how composed skills lead to unauthorized actions and policy violations.

## What to do

No action recorded; read the sources below and the vendor's advisory.

## Latest updates

- 2026-10-27: CoSA: The Conflict of Skills Attack in Real-World LLM Agent Ecosystems (https://arxiv.org/abs/2610.35082)

## Sources

- 2026-10-27 arXiv: CoSA: The Conflict of Skills Attack in Real-World LLM Agent Ecosystems https://arxiv.org/abs/2610.35082

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.