hirundo-io/tool-injection-answered-forget-set

Tool-output prompt-injection forget set (answered targets) A forget set for hardening openai/gpt-oss-20b against indirect prompt injection, meaning attacker instructions that arrive inside tool output (emails, documents, web pages, records). In every row the model completes the user's real task and does not act on the planted instruction. There are no blanket refusals. The set also carries the controls a hardened agent needs so it does not over-refuse: Hard negatives: the user… See the full description on the dataset page: https://huggingface.co/datasets/hirundo-io/tool-injection-answered-forg

<iframe src="https://silytics.si/embed/item/16171ada-5c43-4df4-a8c9-bdc55200500d" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 24 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.