List your company

Agentic Workflow Hijacking Vulnerability Detection on ask-bonk GitHub Action

7.6M.P.AutoDAN-T
0.7362.5184.36.082May 11, 2026

Evaluation results

MethodLinks
AutoDAN-TModel=Claude Haiku 4.52026.057.66
AutoDAN-TModel=Gemini-2.5-Flash2026.054.75.7
JAW-NoPromptProvModel=Claude Sonnet 4.52026.053.35.7
JAW-NoCapProfModel=Claude Sonnet 4.52026.053—
Search-basedModel=Claude Sonnet 4.52026.052.72.7
Search-basedModel=Claude Haiku 4.52026.052.73.5
JAWModel=Claude Sonnet 4.52026.052.35
Search-basedModel=Gemini-2.5-Flash2026.0522
JAW-NoPromptProvModel=Gemini-2.5-Flash2026.051.74
JAW-NoPromptProvModel=Claude Haiku 4.52026.051.33.7
JAW-NoCapProfModel=Claude Haiku 4.52026.051.3—
JAW-NoCapProfModel=Gemini-2.5-Flash2026.051.3—
JAWModel=Claude Haiku 4.52026.0513
JAWModel=Gemini-2.5-Flash2026.0512.7