From 32c4abefc03f1733580e04bdd1082dfd77b1054e Mon Sep 17 00:00:00 2001 From: Rudy Celekli Date: Mon, 19 Oct 2026 15:22:05 -0400 Subject: [PATCH] docs(detection): correct AgentDojo security field meaning in README The README said AgentDojo's `security` field is False when the attack succeeded. AgentDojo injection tasks return security=True when the injection goal is met, and main_benchmark.py copies that value into `is_malicious` (True = attack succeeded), as the line above states. Co-Authored-By: Claude Opus 5.5 --- Detection/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/Detection/README.md b/Detection/README.md index f8a57970..567a3e84 100644 --- a/Detection/README.md +++ b/Detection/README.md @@ -230,7 +230,7 @@ uv run python main_benchmark.py --benchmark agentdojo --concurrent 3 **AgentDojo Ground Truth** (`ground_truth.json`): - `is_malicious`: True if attack succeeded, False if defended -- `security`: AgentDojo's security field (False = attack succeeded) +- `security`: AgentDojo's security field (True = attack succeeded, the injection task's goal was met; copied to `is_malicious`) - `utility`: AgentDojo's utility field (task completion) - `suite`, `user_task`, `injection_task`: Task identifiers