This release includes 1 security fix for security teams reviewing exposed deployments.
Published 2mo
AI & Machine Learning
✓ No known CVEs patched
This release patches 1 known CVE
Topics
agent-orchestration
agent-runtime
ai-coding
claude-code
cli
code-review
+14 more
codex
coding-agent
developer-tools
devtools
evidence
local-first
local-llm
mcp
model-context-protocol
multi-agent
opencode
software-engineering
typescript
workflow-automation
Summary
AI summaryParallel Orchestration, HUD & Cockpit, Goal Management, Doctor & Quality Hardening, Memory & Context, and Branding & Docs all receive new capabilities.
Full changelog
What's New in 0.4.0
Parallel Orchestration (Alpha)
- Coordinator -> Workers -> Reviewer fan-out pattern for multi-agent execution
- Live DAG scheduler with task graph, evidence gates, and role-aware ensemble
- Real-time ETA estimator using completed durations and worker count
- Parallel Live Renderer with 1.5s refresh and alternate-screen support
HUD & Cockpit
- Live Cockpit (
omk hud) with top summary bar, system/Kimi gauges, and responsive layout - Goal-aware display showing run state, TODO sidebar, and changed files
- Chat Cockpit with heartbeat/thinking sampling and banner replacer
Goal Management (Alpha)
omk goal— Codex-style goal management with success criteriaomk verify— Evidence gate verification for completed runs (supports--json)omk summary— Run summary and report generation- Goal compiler converts goals into DAG nodes with bootstrap, coordinator, and verify
Doctor & Quality Hardening
doctor --jsonstructured output for CI consumption- npm 10+ detection using
npm prefix -g(removed deprecatednpm bin -g) - Smoke test hardened to reject unexpected doctor warnings and errors
npm run audit:packagevalidates tarball against allow-list
Agent & Skill Registry
omk agent list/show/create/doctor— 16 built-in agent roles with YAML diagnosticsomk skill pack/install/sync— Curated Kimi skill pack management- Agent roles extend Okabe-compatible base with D-Mail support
Sync (Alpha — with caveats)
omk sync --dry-run,--diff,--rollbackfor safe mutation preview- Note: Global skills symlink rollback is not yet manifest-backed
Spec Kit Integration (Alpha)
omk specify— GitHub Spec Kit CLI bridgeomk dag from-spec— Converttasks.mdinto OMK DAG JSONomk parallel --from-spec— Execute spec-based DAG via parallel orchestrator
Memory & Context
- Local Graph Memory —
.omk/memory/graph-state.jsonwith ontology mindmaps - GraphQL-lite recall via
omk_graph_query - Run state persistence under
.omk/runs/with schema versioning - D-Mail checkpoints + snippets for reusable code blocks
Branding & Docs
- README rebranded as Kimi-native wrapper (not a generic AI tool)
- Alpha warning banner added at the top of README
- Architecture diagrams updated across all 4 language sections
MATURITY.mdandCHANGELOG.mdadded with known limitations
Security & Cleanliness
- Local runtime configs removed from public tracking —
.omk/,.kimi/,.agents/no longer tracked in Git .gitignoreupdated to ignore.omk/entirely- Guard hooks block destructive commands and secret leakage by default
stop-verify.shcomprehensive verification on exit
Known Limitations
parallel,run,verify,goal, andsyncare alpha — expect breaking changesdoctorreports errors when Kimi CLI orjqare missing; use--softfor CI- Chat cockpit telemetry is heartbeat/thinking sampling based, not a full native Kimi event stream
- GitHub Actions matrix (Node 20/22/24 × ubuntu/windows/macos) must pass before this release is considered fully stable
Full Changelog: CHANGELOG.md
Security Fixes
- Local runtime configs removed from public tracking; `.omk/`, `.kimi/`, `.agents/` added to .gitignore, guard hooks block destructive commands and secret leakage
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About OMK
Provider-neutral CLI control plane for coding agents: routes runtimes, scopes MCP, runs DAG workers, and verifies evidence before completion. MIT.
Related context
Related tools
Beta — feedback welcome: [email protected]