Skills
Claude Code 2.1.243 adds a Loops breakdown to /usage and splits prompt-cache TTL between the main conversation and subagents
The 2026-08-25 release adds per-loop run count, total tokens, tokens per run and last-run time to /usage so a runaway /loop task is visible, plus separate promptCacheTtl and subagentPromptCacheTtl settings letting API-key and cloud-provider users hold a 1-hour cache on the main thread while subagents stay at 5 minutes. It also adds a modelPicker setting to curate /model with an ordered labeled list (including Vertex/Bedrock ids), a modelPricing managed setting so /cost and the status line use contracted rates instead of list price, and per-subagent model plus effort level in /tasks. Companion 2.1.245 fixes a startup crash on glibc 2.44 distros (Arch, CachyOS, Fedora Rawhide).
↳ Follow the thread