AI agents aren't safe from prompt injection, and spreadsheets prove it

(shiftmag.dev) by taubek | Sep 6, 2026 | 0 comments on HN