We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
There was an error while loading. Please reload this page.
a309773
v1.0.10: fix jailbreak attack false positives in security scoring - Expand refusal context window from 100 to 200 chars - Skip all forbidden keyword checks for JB attacks when agent refused - Add payload-echo detection to avoid flagging quoted-back attack prompts - Split COMPROMISED_INDICATORS into hard (override refusal) and contextual (respect refusal) tiers - Treat file-path behavioral patterns as benign during refusals - Remove overly generic forbidden keywords (password, credential, exploit, etc.) from 14 JB attack definitions Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
