I've been maintaining a CVE-style database of real, sourced AI agent failures. With 60 catalogued, the patterns are clear enough to write up: security vulnerabilities and destructive actions make up about half, 47% are critical, and the most common root cause is confidence miscalibration - agents acting decisively on unverified assumptions.
This is a companion discussion topic for the original entry at https://www.reddit.com/r/artificial/comments/1uz3771/i_analyzed_60_documented_ai_codingagent_failures/