Repository navigation
Attack benchmark, trained detector and AgentDojo comparison. See the README for results and caveats.
Assets
agentshield-distilbert-v0.2.0.zip: fine-tuned DistilBERT detector (safetensors + tokenizer) and its evaluation metrics. Experimental: 92% recall / 0% FPR on the public test set, but 100% false positives on in-domain agent data. See docs/MODEL_CARD.md.agent_shield-0.2.0-py3-none-any.whl/.tar.gz: the Python package.SHA256SUMS: checksums for the above.