Artifacts for the paper "AgentGuard-DiD: A Generic Runtime Defense-in-Depth Framework for Agentic LLMs" (Kuan-Ju Chen, Po-Wen Chi, NTNU).
Code, scripts, and one-command reproduction:
https://github.com/darren-k-chen/agentguard-did (tag paper-v1).
corpus_manifest.csv — the 66 successful-attack corpus: per-attack, per-method verdicts (L1/L2/L3), level unions, full-stack, qwen success label, and independent… See the full description on the dataset page:
https://huggingface.co/datasets/kuanju/agentguard-did.
