0.2.12

번역 준비 중

이 쪽은 아직 한국어로 옮기지 않았습니다. 아래는 영어 원문입니다.

Behavior changes

• elanous release checklist add warns about an ID used in another release and requires --allow-duplicate-id to proceed; a blocked move now identifies the destination cell.
• CLI and cron outbound sends now register a standalone log-store sink, recording bot identity, kind, source, delivery path, outcome and body length without message text or credentials.
• elanous persona add <name> --from <preset> creates one editable YAML persona; persona edit --set validates fields and returns exit 2 with a one-line reason for unknown keys.
• Harness ask/say launch gates now have an isolated real-CLI test for duplicate blocking, budget blocking, and force overrides, while unknown active-run lookups record their reasons.
• Harness launches warn and proceed when provider usage is unmeasured; only measured exhaustion blocks the launch.
• When a change has more than 30 caller tests, the gate keeps the ones closest to it — tests naming a route on a changed line, then direct importers nearest in the directory tree — instead of the alphabetically first; the overflow still runs in the release-cut sweep.
• Retry a silent release-gate Pod shard once in a fresh Pod before reporting it stalled.
• the release loop now packs the internal ops feed from the release tag (a fresh checkout with the web UI built), so operations hosts receive exactly the released code instead of whatever main held at that moment.
• Harness runs and external agent missions publish fail-soft claimed and done events with run and PR references to the context journal.
• Market review bundles exclude plugin runtime and graph-run state while retaining the private-content leak gate for authored files.
• Persist ordered agent-mission PTY decisions to logs.db so the PWA inside scene can replay them.
• when a harness change edits a test, the gate again also runs the changed sources' co-located and related tests (a narrowing that ran only the edited test is reverted).

Plus 81 internal changes.