release aidev PRERELEASE SIG 3/5

OpenClaw 2026.6.7-beta.1: Kimi K2.7 Code support, guarded Skill Workshop writes

Beta prerelease adding the Kimi K2.7 Code model to the provider catalog with reasoning content preserved across tool turns. Skill Workshop support-file writes are now constrained to guarded lifecycle writes rather than unbounded filesystem updates, plus a batch of channel and provider-session recovery fixes.

PUBLISHED2026-06-13
OBSERVED2026-08-11
AGE1mo
SOURCES1
  • Providers: Kimi K2.7 Code added; replay preserves its reasoning content across tool turns
  • Skills: Skill Workshop support-file targets only writable through guarded lifecycle writes; package publishing moved to current ClawHub plugin checks
  • Provider sessions: SecretRef-backed profile auth, configured DeepSeek transport inheritance, static model fallback, Anthropic thinking replay, and Codex prompt memory all recover instead of leaving a turn unusable
  • CLI-backed /btw fallback now fails closed on unsupported paths
  • Channels: Telegram polling and preview failures recover instead of ending the stream; Slack final replies stay in the transcript; top-level outbound image parameters preserved
  • Operations: cron status reports the SQLite storage path instead of legacy jobs.json; disabled-heartbeat one-shot retries preserved; Linux service auto-updates handed off; daemon status stays readable on unsupported services
  • Anthropic Vertex: stops re-marking cache_control on transport-budgeted payloads

COMMUNITY

No curated reactions recorded for this event. Facts and takes are kept in separate layers — community context is added by hand, never blended into the record above.