Security
assume hostile input
Security
Prompt injection and tool permissions: a practical defence
Why untrusted text can steer your agent, and the permission rules that limit the damage.
Rajath Kumar K S
Topicassume hostile input
Security
Prompt injection, tool permissions, secrets and safe defaults for AI systems.
1 article in this topic · All topics
Featured article in Security
Explore other topics
- AI Agents2 articles
- LLM Engineering1 article
- PythonComing soon
- FastAPI & Backend1 article
- Evaluation & Testing1 article
- Deployment & MLOpsComing soon
- Applied AIComing soon
- Tutorials & Case StudiesComing soon