search
prompt injection
Trends
- 1AI agents raise questions over control and authorization▼AI agents are getting more powerful but who controls their power? When an AI can read emails, access files and call APIs
Security commentators are warning that as AI agents gain the ability to read emails, access files and call APIs, a hidden prompt injection can escalate from producing a bad answer to causing real damage. The argument being made is that capability does not equal authorization: least-privilege access, human approval and continuous monitoring must be built in rather than treated as optional safeguards.
Repos
- kerpopule/hermes-jev-skills Jev-powered model routing, memory, compaction, skill selection, computer and browser use for Hermes agents (also Claude
- rudratoshs/buried-injections 🛡️ Regex catches 0%, Meta's Prompt Guard 2 catches 1% of 629 realistic AgentDojo injection attacks when they'r