Claude Code v2.1.210 Hardens the Agent Tool Against Indirect Prompt Injection and Stops 'ultracode' From Firing on Webhook Input

Claude Code Changelog·high signal

v2.1.210 fixes the `ultracode` keyword opt-in firing on non-human-originated input such as webhooks and relayed PR comments, and separately hardens the Agent tool against indirect prompt injection. The ultracode fix is the notable one: a keyword that escalates orchestration scale was triggerable by anyone who could get text into a relayed PR comment — a cost-and-blast-radius escalation from untrusted input.