AI BEAT / SEED COMMUNITY
Seed Collections
Search and filter four risk collections, then open examples and loading guides.
Search terms appear in the URL. Avoid sensitive information.
Content Safety
Output policy, harmful content, and refusal checks.
View example & tutorialInstruction & Prompt Injection
Trusted instruction boundaries and prompt-injection surfaces.
View example & tutorialPermission & Tool Safety
Sensitive tool calls, unauthorized actions, and confirmation boundaries.
View example & tutorialPrivacy & Data Safety
Sensitive data leakage and privacy flows.
View example & tutorialNo risk collections match these filters.
Clear a filter or try another term. Unclassified AgentDojo scenes stay off this list until they carry a risk tag.