In sandboxed runs, Gemini CLI declared intent to run malicious preflight commands in 96.1% of cases and Qwen Code in 74.0%, while only one run showed verified execution.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SE 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Towards a Risk Assessment of Malicious Skill Files in Coding Agents
In sandboxed runs, Gemini CLI declared intent to run malicious preflight commands in 96.1% of cases and Qwen Code in 74.0%, while only one run showed verified execution.