v0.4.0

What's Changed

• feat(skills): agent-agnostic skill architecture with Kimi support by @caoergou in #16
• feat(skills): enhance verify with change traps and add manual runner skill by @caoergou in #24
• feat(skills): split verify into references and align skills with runner by @caoergou in #29
• fix(engine): include iteration in done_signal event for reporter by @caoergou in #26
• fix(engine): sync existing .owloop/ into worktree instead of skipping copy by @caoergou in #28
• feat: extend owloop go with all owloop run runtime options by @caoergou in #27
• feat(engine): per-session worktree branches with --resume by @caoergou in #30
• feat(spec_queue): respect spec dependency order in engine scheduling by @caoergou in #31
• fix(engine): strengthen final done signal instruction in BUILD_PROMPT by @caoergou in #38
• docs(skills): enhance owloop-verify with workflow, severity levels, and reference checklists by @caoergou in #39
• feat(engine): session/resume semantics with budget carry-over by @caoergou in #40
• feat(spec_generator): inject linter rules into generation prompt and auto-retry on lint errors by @caoergou in #41
• feat(cli): add --no-tui/--plain flag to owloop run by @caoergou in #44
• Batch: structured event log, dry-run, token tracking, and status bold-marker fix by @caoergou in #48
• feat(engine): structured JSONL event log for debugging and reporting by @caoergou in #42
• feat(engine): add --dry-run/--one-shot mode to owloop run by @caoergou in #43
• feat(engine): real-time token tracking with per-iteration budget cap by @caoergou in #46
• fix(cli): recognize bold Status: markers in owloop status by @caoergou in #47

New Contributors

• @caoergou made their first contribution in #16

Full Changelog: v0.3.1...v0.4.0