~/lieflab-os $ git log --oneline_
v0.0.62026-06-23(analysis)
project analysis + resilient loop
+analysis field added to project YAML schema + CRUD endpoints
+home right panel restored: Claude window + active-agent panel
+agent_loop stops silently skipping issues on usage limit / empty output
v0.0.52026-06-17(hardening)
the os enforces its own rules
+structural hardening — agent OS invariants mechanically enforced across all five areas
+self-QA loops + automated gating for agent autonomy
+scheduled batch runner for active projects + run-backlog <file> CLI
+per-project todo lists managed as markdown files
+CTO pre-execution review gate for large issues · PRD/issue breakdown rule
v0.0.42026-06-16(workspace)
the cockpit comes together
+embedded terminal panel — xterm.js + PTY in the right column
+right-side Claude chat panel for local interaction
+needs-human signals batched and deferred to natural breakpoints
+per-item execution status written to state/active_tasks.md
v0.0.32026-06-13(orchestration)
many agents, one queue
+parallel multi-agent dispatch for tasks needing multiple specialists
+type:recurring issues — reset to status:todo after merge
+Today's Report — project markdown reports as clickable date history
v0.0.22026-06-11(agents)
agent orchestration online
+agent_loop pipeline — branch › test › commit › PR › squash-merge › auto-close
+parallel fan-out for non-conflicting tasks + per-project locking
+gitbot writes full issue title+body from a one-line idea
+GITBOT_TOKEN — agent issues/comments post as the bot account
+international label taxonomy drives the pipeline (type / priority / size / scope)
v0.0.12026-06-10(seed)
first skeleton + the single dashboard
+repo + CLAUDE.md / plan.md / engineering-method protocol
+21 persona prompts + orchestrator core (Python, minimal deps)
+FastAPI single-page dashboard — overview · projects · KPIs · agents · docs
+state/active_tasks.md = single source of truth, 3-rule handoff protocol
# github.com/RosieWikiLab/lief-os · MIT License
