Dataset · Hugging Face

ZJUICSR/DeepTrap

Name: DeepTrap
Creator: ZJUICSR
License: mit
Keywords: task_categories:text-generation, license:mit, size_categories:n<1K, format:json, modality:text, library:datasets, library:pandas, library:polars, library:mlcroissant, arxiv:2605.11047, region:us, benchmark, agent-safety, security, prompt-injection

DeepTrap Security benchmarks for OpenClaw agents DeepTrap is a benchmark suite for evaluating how OpenClaw agents behave when benign user tasks are mixed with malicious workspace context, injected skills, misleading tool metadata, unsafe command paths, and encoded payloads. The public repository is intentionally minimal: it contains the runner, the bundled replay tasks, and the scoring code needed to evaluate OpenClaw models. It does not include the private task-generation or… See the full description on the dataset page: https://huggingface.co/datasets/ZJUICSR/DeepTrap.

View on Hugging FaceReport an issue

Author / organization	ZJUICSR
License (as declared)	mit
Downloads	163 (as reported by Hugging Face)
Likes	0
Created	16 May 2026
Last updated at source	16 May 2026

benchmarkagent-safetysecurityprompt-injectionllm-agentsopenclaw

Embed this entry

<iframe src="https://silytics.si/embed/item/c8496510-236e-4d7f-8fdd-de5f70624915" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Free, no tracking. More about feeds and embeds.

More from ZJUICSR

ZJUICSR/ActBenchDataset

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 16 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.