Tag: social-engineering

Designing agent workflows to limit prompt injection and social‑engineering risks

Build agent workflows that separate fetch, intent extraction, decision, and action; record provenance, gate capabilities, and require human confirmation to curb prompt-injection risks.