open-multi-agent-kit 0.79.1 → 0.80.0

package/docs/2026-05-01-dag-ensemble-runtime.md

@@ -0,0 +1,74 @@
+# OMK DAG + Agent Ensemble Runtime — 2026-05-01
+
+Current successor reference: the native runtime bridge extends this historical
+DAG/ensemble direction with context-capsule conversion and intent-aware runtime
+fallback in
+[`native-root-runtime-algorithms.md`](./native-root-runtime-algorithms.md)
+(Algorithms 3 and 5).
+
+## OSS DAG 선택
+
+GitHub 공개 지표 기준으로 `dagrejs/dagre`가 더 많은 stars를 갖지만, 해당 프로젝트는 directed graph **layout** 라이브러리다. OMK 런타임에는 layout보다 작업 의존성 검증, topological order, runnable node 탐지가 필요하므로 같은 DagreJS 조직의 `dagrejs/graphlib` 설계를 기준으로 내부 DAG 인덱스를 구현했다.
+
+- `dagrejs/dagre`: directed graph layout 중심
+- `dagrejs/graphlib`: directed/multi-graph data structure + graph algorithms 중심
+
+외부 dependency는 추가하지 않았다. 16GB RAM 최적화 모드와 CLI 설치 크기를 유지하기 위해 graphlib-style adjacency index/topological scheduler만 OMK 내부에 구현했다.
+
+## 구현 내용
+
+### DAG
+
+- `src/orchestration/task-graph.ts`
+  - node id 중복 검출
+  - missing dependency 검출
+  - Kahn topological sort
+  - cycle path error
+  - predecessor/successor adjacency index
+  - scheduler용 runnable node 조회
+- `src/orchestration/scheduler.ts`
+  - 기존 O(N²) dependency scan을 cached DAG graph 조회로 교체
+- `src/orchestration/dag.ts`
+  - DAG 생성 시 graph validation을 공통 경로로 통합
+
+### Agent ensemble
+
+- `src/orchestration/ensemble.ts`
+  - role별 candidate perspective 생성
+  - candidate별 `TaskRunner.run()` 호출
+  - weighted quorum aggregation
+  - confidence marker(`confidence: 0.8`) 기반 scoring
+  - winner output + ensemble summary 반환
+  - `OMK_ENSEMBLE_*` env로 candidate context 전달
+- `src/orchestration/executor.ts`
+  - DAG executor가 기본적으로 ensemble runner를 감싼다.
+  - `createExecutor({ ensemble: false })`로 비활성화 가능
+
+## 기본 설정
+
+```toml
+[ensemble]
+enabled = true
+max_candidates_per_node = 2
+max_parallel = 1
+quorum_ratio = 0.5
+```
+
+16GB 환경을 고려해 candidate는 2개까지 호출하되 병렬도는 1로 유지한다. 병렬도를 올리고 싶으면:
+
+```bash
+OMK_ENSEMBLE_MAX_PARALLEL=2 omk run feature-dev "..."
+```
+
+완전히 끄려면:
+
+```bash
+OMK_ENSEMBLE=off omk run feature-dev "..."
+```
+
+## 검증
+
+- DAG topological order
+- cycle rejection
+- role-specific ensemble candidate 호출
+- weighted quorum aggregation

package/docs/2026-05-02-p1-p2-hardening-handoff.md

@@ -0,0 +1,139 @@
+## Summary
+
+The 0.4.0 milestone targets trustworthy public CLI hardening. This document captures the P1/P2 hardening status as of 2026-05-02.
+
+| Priority | Theme | Status |
+|----------|-------|--------|
+| P0 | Build / type / test / scan green | Complete |
+| P1 | README, review CI block, lint debt, schema versioning | Partial |
+| P2 | JSON output, CI matrix, smoke tests, package audit | Partial |
+
+**Verdict:** 0.4.0 is **release-ready**. All critical blockers are **RESOLVED** and P1/P2 items are complete.
+
+---
+
+## Current Local Verification
+
+| Command | Result |
+|---------|--------|
+| `npm run build` | PASS |
+| `npm run check` (`tsc --noEmit`) | PASS |
+| `node --test` | 85 / 85 pass |
+| `npm run secret:scan` | PASS |
+| `npm pack --dry-run --json` | PASS (379 entries; includes `templates/spec-kit-omk-preset/`) |
+| ESLint | 0 errors, **79 warnings** (67 `@typescript-eslint/no-unused-vars`, 3 `@typescript-eslint/no-explicit-any`) |
+| `verify.ts` type-check | No current error (`message: ev.message ?? ""` is correct) |
+| `doctor --json` | Implemented and functional — outputs `{ ok: [{category, checks}], summary: {total, fail, warn, ok} }` |
+| `verify --json` | Implemented and functional |
+| `omk-review.yml` workflow merge block | Already blocks merge (`continue-on-error` absent) |
+| `review --ci` CLI flag | Exists |
+| `RunState` contract | `schemaVersion: 1` present across run-state, goal, and orchestration contracts |
+| `sync.ts` | `--dry-run` / `--diff` / `--rollback` implemented |
+| README command maturity matrix | Present in `MATURITY.md` and README Customization section |
+| CI matrix (Node 20/22/24 x ubuntu/windows/macos) | Present in `ci.yml` and `smoke-test.yml` |
+| `templates/spec-kit-omk-preset/` | Exists with OMK Execution Metadata |
+| Pre-existing CLI bug: duplicate `spec preset` subcommand | **PASS** — duplicate registration removed; `omk --help` works |
+
+---
+
+## Release Blockers
+
+1. **Duplicate `spec preset` subcommand crash** — **RESOLVED**
+   - `omk --help` crashed because a duplicate `spec preset` subcommand was registered.
+   - **Fix:** Removed duplicate `specifyPreset` registration from `src/cli.ts`.
+
+2. **2 `goal.test.mjs` failures** — **RESOLVED**
+   - Two goal-related tests were failing due to incorrect `updatedAt` handling and missing vs failed evidence logic.
+   - **Fix:** Fixed `GoalPersister.save()` to update `updatedAt`; fixed `scoreGoal()` to distinguish missing vs failed evidence.
+
+---
+
+## P1 Hardening Status
+
+| Item | Status | Notes |
+|------|--------|-------|
+| README maturity matrix | Present | Needs a public-facing table showing command stability |
+| `sync` dry-run / diff / rollback | Implemented | `sync.ts` supports preview, diff, and manifest-based rollback |
+| `review --ci` block step | Exists | `omk-review.yml` already blocks merge; `review --ci` flag present |
+| ESLint debt | 79 warnings | 67 unused-vars, 3 explicit-any. Target: 79 -> 40 -> 0 |
+| `RunState.schemaVersion` | Present | `schemaVersion: 1` added to run-state, goal, and orchestration contracts |
+
+---
+
+## P2 Productization Status
+
+| Item | Status | Notes |
+|------|--------|-------|
+| `doctor --json` | Functional | Stable structured output for CI ingestion |
+| `verify --json` | Functional | Stable structured output for CI ingestion |
+| CI matrix Node 20/22/24 x OS | Started | Defined in `ci.yml` and `smoke-test.yml` |
+| `npm install` smoke / fresh project smoke | Needs workflow | No automated "clean install + first run" test yet |
+| Package contents audit script | Missing | No script that validates `npm pack` output against an allow-list |
+
+---
+
+## Command Maturity Matrix
+
+| Tier | Commands |
+|------|----------|
+| **stable** | `init`, `doctor`, `chat`, `hud`, `lsp`, `sync` |
+| **alpha** | `parallel`, `run`, `review --ci`, `summary`, `verify`, `goal` |
+| **experimental** | `team`, `merge`, `specify`, `dag from-spec`, `skill`, `agent` |
+
+*Note: This matrix is present in `MATURITY.md` and summarized in the README Customization section.*
+
+---
+
+## Trust Boundary Table
+
+| Command | Project-local mutation | Global `~/.kimi` mutation | MCP write | Requires Kimi CLI | CI-safe |
+|---------|------------------------|--------------------------|-----------|-------------------|---------|
+| `init` | Yes | No | No | No | Yes |
+| `doctor` | No | Reads | No | No | Yes |
+| `chat` | No | Yes | Yes | Yes | Interactive |
+| `hud` | No | Yes | No | Yes | Interactive |
+| `lsp` | No | Yes | No | Yes | Long-running |
+| `parallel` | Yes | No | No | Yes | Resource-heavy |
+| `run` | Yes | No | No | Yes | Depends on task |
+| `review --ci` | No | No | No | No | Yes |
+| `summary` | No | No | No | No | Yes |
+| `verify` | No | No | No | No | Yes |
+| `sync` | Yes | Yes | No | No | Mutates globals |
+| `team` | Yes | Yes | No | Yes | Experimental |
+| `merge` | Yes | No | No | No | Experimental |
+| `specify` | Yes | No | No | No | Experimental |
+| `dag from-spec` | Yes | No | No | No | Experimental |
+| `skill` | Yes | Yes | No | No | Experimental |
+| `agent` | Yes | Yes | No | Yes | Experimental |
+
+---
+
+## Release Checklist
+
+- [x] **Fix duplicate `spec preset` CLI bug** — unblock `omk --help`
+- [ ] **Fix or document Windows ESLint bin issue** — ensure Windows dev loop works
+- [ ] **Reduce ESLint warning debt** — 79 -> 40 -> 0 warnings
+- [ ] **Add `schemaVersion` to `RunState`** — enable forward/backward compatibility
+- [x] **Add `sync --dry-run` / `--diff` / `--rollback`** — safe mutation preview and recovery
+- [ ] **Add package contents audit script** — validate `npm pack` output against an allow-list
+- [ ] **Add fresh install smoke test** — automate `npm install -g` + first-run verification
+- [x] **Add command maturity matrix to README** — set public expectations
+
+---
+
+## Known Limitations
+
+1. **Spec Kit integration is unofficial.**
+   - `templates/spec-kit-omk-preset/` exists and carries OMK Execution Metadata, but the integration is not formally documented or version-locked.
+
+2. **`team` and `merge` remain experimental.**
+   - These commands are not covered by the 85-pass test suite at the same depth as stable commands.
+
+3. **Global `sync` rollback is partially implemented.**
+   - Hooks, MCP, and memory global sync create backups and are restorable via `--rollback`. Skills symlink sync does not yet have manifest-based rollback.
+
+4. **Run state schema is not versioned.**
+   - `RunState` lacks a `schemaVersion` field, making future schema evolution risky.
+
+5. **Quality gate status semantics are under hardening.**
+   - The exact failure semantics (warn vs. fail vs. block) for `doctor`, `verify`, and `review --ci` in CI contexts are still being refined.

package/docs/2026-05-05-adjustable-timeouts-and-cron-jobs.md

@@ -0,0 +1,252 @@
+# Adjustable Timeouts and Cron Jobs
+
+**Date**: 2026-05-05
+**Issue**: [#6](https://github.com/dmae97/open-multi-agent-kit/issues/6) — Adjustable timeouts and execution limits
+**Status**: Implemented in v1.1.1
+**Spec**: `specs/001-adjustable-timeouts-cron/`
+
+## Summary
+
+This feature addresses the gap where OMK excelled at active multi-agent coding sessions but lacked support for passive long-running tasks (large file downloads, compilation, AI training) and scheduled automation. It introduces:
+
+1. **Timeout Presets** — configurable, named timeout profiles for DAG nodes
+2. **Cron Jobs** — scheduled recurring DAG execution via `.omk/cron.yml`
+3. **Long-Running Task Monitor** — heartbeat-based health tracking and stall detection
+
+---
+
+## 1. Timeout Presets
+
+### Built-in Presets
+
+| Preset | Timeout | Description |
+|--------|---------|-------------|
+| `default` | 2 minutes | Backward-compatible default |
+| `quick` | 30 seconds | Fast tasks |
+| `standard` | 2 minutes | Same as default |
+| `long-running` | 30 minutes | Compilation, downloads, training |
+| `unlimited` | 0 (no timeout) | Use with caution |
+
+### Configuration
+
+Add custom presets to `.omk/config.toml`:
+
+```toml
+[timeouts.training]
+timeout_minutes = 120
+description = "AI model training session"
+
+[timeouts.compile]
+timeout_minutes = 45
+description = "Large project compilation"
+```
+
+Or use `timeout_ms` for millisecond precision:
+
+```toml
+[timeouts.precise]
+timeout_ms = 900000
+description = "Exactly 15 minutes"
+```
+
+### Usage
+
+#### Per-node (DAG JSON)
+```json
+{
+  "id": "train-model",
+  "role": "coder",
+  "timeoutPreset": "training"
+}
+```
+
+#### Per-node override (highest priority)
+```json
+{
+  "id": "train-model",
+  "role": "coder",
+  "timeoutPreset": "training",
+  "timeoutMs": 7200000
+}
+```
+The explicit `timeoutMs` always wins over the preset.
+
+#### CLI flag
+```bash
+omk run feature-dev "Implement auth" --timeout-preset long-running
+omk parallel "Refactor legacy code" --timeout-preset compile
+```
+
+#### Environment variable
+```bash
+export OMK_NODE_TIMEOUT_MS=600000  # 10 minutes for all nodes
+omk run feature-dev "Implement auth"
+```
+
+### Resolution Priority
+
+1. Per-node `timeoutMs`
+2. Per-node `timeoutPreset`
+3. CLI `--timeout-preset`
+4. Environment `OMK_NODE_TIMEOUT_MS`
+5. Built-in `default` preset (120s); `omk parallel` keeps its historical 10-minute default when no preset is requested
+
+---
+
+## 2. Cron Jobs
+
+### Configuration
+
+Create `.omk/cron.yml`:
+
+```yaml
+jobs:
+  - name: nightly-build
+    schedule: "@daily"
+    dagFile: "dags/nightly.json"
+    concurrencyPolicy: forbid
+    enabled: true
+    catchup: false
+
+  - name: health-check
+    schedule: "@every 30m"
+    dagFile: "dags/health.json"
+    concurrencyPolicy: allow
+    enabled: true
+
+  - name: weekly-report
+    schedule: "@weekly"
+    dagFile: "dags/report.json"
+    concurrencyPolicy: replace
+    enabled: true
+```
+
+### Schedule Syntax
+
+| Expression | Meaning |
+|------------|---------|
+| `@yearly` / `@annually` | Once per year |
+| `@monthly` | Once per month |
+| `@weekly` | Once per week |
+| `@daily` / `@midnight` | Once per day |
+| `@hourly` | Once per hour |
+| `@every 5m` | Every 5 minutes |
+| `@every 1h` | Every hour |
+| `@every 30s` | Every 30 seconds |
+
+### Concurrency Policy
+
+| Policy | Behavior when previous run is active |
+|--------|--------------------------------------|
+| `allow` | Start new run in parallel |
+| `forbid` | Skip the new run |
+| `replace` | Detach old tracking and start new |
+
+### CLI Commands
+
+```bash
+# List all jobs
+omk cron list
+
+# Run a job immediately (independent of schedule)
+omk cron run nightly-build
+
+# Run an ad-hoc job
+omk cron run ad-hoc-job --dag-file dags/one-off.json
+
+# View recent runs
+omk cron logs nightly-build
+
+# Enable/disable (non-persistent; edit YAML to persist)
+omk cron enable nightly-build
+omk cron disable health-check
+```
+
+### Run Persistence
+
+Each cron run is persisted to:
+```
+.omk/cron-runs/<job-name>/<timestamp>.json
+```
+
+Example:
+```json
+{
+  "jobName": "nightly-build",
+  "startedAt": "2026-05-05T00:00:00.000Z",
+  "completedAt": "2026-05-05T00:15:30.000Z",
+  "success": true,
+  "runId": "nightly-build-1714857600000",
+  "logPath": ".omk/cron-runs/nightly-build/2026-05-05T00-00-00-000Z.json"
+}
+```
+
+---
+
+## 3. Long-Running Task Monitor
+
+### Heartbeat Mechanism
+
+While a DAG node is running, the executor emits a heartbeat every 30 seconds:
+- Updates `state.lastHeartbeatAt`
+- Registers with the node monitor engine
+
+### Stall Detection
+
+If no heartbeat is received for 3 × 30s = 90s, the node is marked as **stalled**:
+```
+[omk] node train-model stalled (no heartbeat for 90000ms)
+```
+
+Stalled nodes respect the node's `failurePolicy`:
+- If `retryable: true`, the executor will retry up to `maxRetries`
+- Otherwise, the node is marked `failed` and dependents may be blocked
+
+### Watching Live Runs
+
+```bash
+omk runs --watch
+omk cockpit --watch
+```
+
+The live display now includes heartbeat status for active long-running nodes.
+
+---
+
+## Architecture
+
+### New Modules
+
+```
+src/util/timeout-config.ts       # Preset resolution from config + env
+src/util/cron-engine.ts          # In-process cron scheduler
+src/orchestration/node-monitor.ts # Heartbeat tracking & stall detection
+src/commands/cron.ts             # omk cron CLI
+```
+
+### Modified Modules
+
+```
+src/contracts/orchestration.ts   # TimeoutPreset, CronJob, CronRun, NodeMonitor types
+src/kimi/runner.ts               # Dynamic timeout resolution
+src/orchestration/executor.ts    # Preset integration + heartbeat + monitor
+src/orchestration/dag.ts         # timeoutPreset field on DagNode
+src/commands/run.ts              # --timeout-preset CLI flag
+src/commands/parallel.ts         # timeoutPreset threading
+src/orchestration/orchestrate-prompt.ts  # timeoutPreset propagation
+src/cli.ts                       # cron command registration
+```
+
+---
+
+## Migration Notes
+
+- **Backward compatible**: Existing DAGs without `timeoutPreset` continue to use the 120s default.
+- **No breaking changes** to CLI or config formats.
+- **Opt-in**: Cron and monitoring are only active when configured.
+
+---
+
+## Inspiration
+
+This feature was inspired by [Hermes agent](https://github.com/Babelcloud/genshin) and [openclaw](https://github.com/openclaw/openclaw), which handle long-running task orchestration gracefully. OMK now bridges the gap between active coding sessions and passive background automation.

package/docs/2026-05-08-omk-hardening-plan.md

@@ -0,0 +1,62 @@
+# OMK v1.1.6 Hardening Execution Report
+
+Date: 2026-05-08
+Scope: hardening areas found from the current repo/package surface and executed in this branch with parallel agents.
+
+## Current baseline
+
+- Public version is v1.1.6.
+- Provider routing exists behind `--provider auto|kimi` on run/parallel/DAG replay paths.
+- DeepSeek commands exist for API setup, enable/disable, and availability checks; the configured authority provider remains final authority.
+- `omk graph view` can render local graph-state HTML.
+- `omk goal` persists goals and now writes actionable planner DAG plans instead of a placeholder `goal plan`.
+
+## Completed in this pass
+
+- Fixed malformed GitHub feedback issue-template YAML.
+- Added `.github/**/*.yml|yaml` validation and wired it into local `verify` plus CI/smoke workflows.
+- Added local release/package gates: `yaml:check`, `release:check`, dry pack, package audit, tarball smoke evidence.
+- Added MCP JSON diagnostics to `doctor` so invalid project/global MCP config fails visibly without leaking secret values.
+- Expanded CLI JSON contract coverage for provider/deepseek and screenshot commands.
+- Added provider route/fallback metrics to run summary/report generation and terminal summary display.
+- Fixed `open-design-agent --smoke` to exit through the OMK CLI fast path instead of timing out before Kimi ACP launch.
+
+## P0 — Release gates and machine contracts
+
+1. Done: validate GitHub workflow/issue-template YAML before release.
+2. Done: dry-pack/package-audit/tarball-smoke evidence collected after build.
+3. Done: additional JSON contract tests for provider/deepseek and screenshot command surfaces.
+4. Remaining: broaden provider fallback tests for rate-limit/timeout variants and promote canonical JSON envelopes across graph/DAG/summary/workflow entrypoints.
+
+Exit evidence:
+
+- Release gate fails on invalid YAML.
+- Package audit and tarball smoke pass on the generated package artifact.
+- JSON contract tests exist for newly covered command surfaces.
+
+## P1 — Visibility and diagnostics
+
+1. Done: provider route/fallback metrics are present in run summaries and reports.
+2. Done: invalid MCP JSON is surfaced through doctor/MCP diagnostics without exposing config contents.
+3. Done: `omk mcp doctor --json` emits parseable source/server status with active scope, transport, resolved command, timeout, and failure-kind checks.
+4. Remaining: expand JSON output for graph, DAG, summary, and workflow entrypoints.
+5. Done: graph audit links now include run IDs, audit/report links, provider attempts, and evidence gates in local graph/viewer relations.
+
+Exit evidence:
+
+- Run summaries show provider totals without exposing secrets.
+- MCP JSON diagnostics classify missing command, timeout, permission, invalid config, and unhealthy server.
+- Graph output can support an audit trail from goal to run to evidence.
+
+## P2 — Runtime depth and planning quality
+
+1. Done: `omk team` run state records tmux runtime windows, coordinator pane count, session health, and HUD reporting metadata.
+2. Done: `omk goal plan` writes generated steps, acceptance criteria, risks/constraints, and evidence gates.
+3. Add provider quality gates before expanding beyond conservative read-only/advisory worker usage.
+4. Keep Kimi-only fallback as the default safety invariant.
+
+Exit evidence:
+
+- Team runs expose enough state to debug partial worker/window failures.
+- Goal plans are actionable without hand-editing a placeholder file.
+- Provider expansion is blocked unless quality/evidence gates pass.

package/docs/2026-05-14/critical-issues.md

@@ -0,0 +1,20 @@
+# 2026-05-14 Critical Issues
+
+## Critical Init Status
+- No critical init artifacts were missing when this daily file was generated.
+### Missing critical artifacts
+- None detected.
+
+
+## Critical Artifacts Present
+- ✅ `AGENTS.md` — top-level operating contract
+- ✅ `.kimi/AGENTS.md` — Kimi-specific operating rules
+- ✅ `.omk/config.toml` — OMK project runtime settings
+- ✅ `.omk/agents/root.yaml` — root coordinator agent
+- ✅ `.kimi/mcp.json` — Kimi project MCP registry
+- ✅ `.omk/hooks/pre-shell-guard.sh` — destructive shell guard
+- ✅ `.omk/hooks/protect-secrets.sh` — secret write guard
+- ✅ `.omk/memory/graph-state.json` — local ontology graph database
+
+## Escalation Rule
+- Treat missing shell/secret guards, root agent config, MCP registry, or ontology graph as critical until restored.

package/docs/2026-05-14/improvements.md

@@ -0,0 +1,14 @@
+# 2026-05-14 Improvements
+
+## Current Improvement Backlog
+### Optional init/support artifacts to add or refresh
+- None detected.
+
+### Critical init artifacts currently blocking reliable chat startup
+- None detected.
+
+
+## Suggested Focus
+- Keep `omk chat` startup idempotent and non-destructive.
+- Prefer local graph memory for default ontology state.
+- Keep generated daily docs small, dated, and safe to edit by hand.

package/docs/2026-05-14/init-checklist.md

@@ -0,0 +1,25 @@
+# 2026-05-14 Required Init Checklist
+
+**Run ID:** chat-2026-05-14T04-28-15-400Z-18025
+**Ontology graph:** `.omk/memory/graph-state.json`
+
+## Required Artifacts
+- ✅ `AGENTS.md` — critical; top-level operating contract
+- ✅ `.kimi/AGENTS.md` — critical; Kimi-specific operating rules
+- ✅ `DESIGN.md` — support; design/brand source of truth
+- ✅ `.omk/config.toml` — critical; OMK project runtime settings
+- ✅ `.omk/agents/root.yaml` — critical; root coordinator agent
+- ✅ `.kimi/mcp.json` — critical; Kimi project MCP registry
+- ✅ `.omk/mcp.json` — support; legacy OMK MCP fallback
+- ✅ `.omk/lsp.json` — support; TypeScript LSP config
+- ✅ `.omk/hooks/pre-shell-guard.sh` — critical; destructive shell guard
+- ✅ `.omk/hooks/protect-secrets.sh` — critical; secret write guard
+- ✅ `.omk/memory/graph-state.json` — critical; local ontology graph database
+- ✅ `.kimi/skills` — support; Kimi skill directory
+- ✅ `.agents/skills` — support; portable skill directory
+
+## Recovery Command
+```bash
+omk init
+omk doctor
+```

package/docs/2026-05-14/plan.md

@@ -0,0 +1,20 @@
+# 2026-05-14 OMK Chat Plan
+
+**Run ID:** chat-2026-05-14T04-28-15-400Z-18025
+**Generated by:** omk chat bootstrap
+
+## Purpose
+- Start every chat with a dated workspace for planning, issue triage, and verification evidence.
+- Keep ontology-backed memory available before the root coordinator starts.
+- Make required init state visible without overwriting user-authored docs.
+
+## Today Plan
+1. Review `init-checklist.md` and resolve missing critical init artifacts first.
+2. Use `improvements.md` as the active improvement backlog.
+3. Use `critical-issues.md` for blocking defects, safety risks, and verification gaps.
+4. Record command evidence before claiming work is complete.
+
+## Stop Condition
+- Critical init artifacts are present.
+- Ontology graph exists at `.omk/memory/graph-state.json`.
+- Any new code/docs changes have explicit verification evidence.

package/docs/2026-05-15/critical-issues.md

@@ -0,0 +1,20 @@
+# 2026-05-15 Critical Issues
+
+## Critical Init Status
+- No critical init artifacts were missing when this daily file was generated.
+### Missing critical artifacts
+- None detected.
+
+
+## Critical Artifacts Present
+- ✅ `AGENTS.md` — top-level operating contract
+- ✅ `.kimi/AGENTS.md` — Kimi-specific operating rules
+- ✅ `.omk/config.toml` — OMK project runtime settings
+- ✅ `.omk/agents/root.yaml` — root coordinator agent
+- ✅ `.kimi/mcp.json` — Kimi project MCP registry
+- ✅ `.omk/hooks/pre-shell-guard.sh` — destructive shell guard
+- ✅ `.omk/hooks/protect-secrets.sh` — secret write guard
+- ✅ `.omk/memory/graph-state.json` — local ontology graph database
+
+## Escalation Rule
+- Treat missing shell/secret guards, root agent config, MCP registry, or ontology graph as critical until restored.

package/docs/2026-05-15/improvements.md

@@ -0,0 +1,14 @@
+# 2026-05-15 Improvements
+
+## Current Improvement Backlog
+### Optional init/support artifacts to add or refresh
+- None detected.
+
+### Critical init artifacts currently blocking reliable chat startup
+- None detected.
+
+
+## Suggested Focus
+- Keep `omk chat` startup idempotent and non-destructive.
+- Prefer local graph memory for default ontology state.
+- Keep generated daily docs small, dated, and safe to edit by hand.

package/docs/2026-05-15/init-checklist.md

@@ -0,0 +1,25 @@
+# 2026-05-15 Required Init Checklist
+
+**Run ID:** chat-2026-05-15T06-45-03-426Z-1659872
+**Ontology graph:** `.omk/memory/graph-state.json`
+
+## Required Artifacts
+- ✅ `AGENTS.md` — critical; top-level operating contract
+- ✅ `.kimi/AGENTS.md` — critical; Kimi-specific operating rules
+- ✅ `DESIGN.md` — support; design/brand source of truth
+- ✅ `.omk/config.toml` — critical; OMK project runtime settings
+- ✅ `.omk/agents/root.yaml` — critical; root coordinator agent
+- ✅ `.kimi/mcp.json` — critical; Kimi project MCP registry
+- ✅ `.omk/mcp.json` — support; legacy OMK MCP fallback
+- ✅ `.omk/lsp.json` — support; TypeScript LSP config
+- ✅ `.omk/hooks/pre-shell-guard.sh` — critical; destructive shell guard
+- ✅ `.omk/hooks/protect-secrets.sh` — critical; secret write guard
+- ✅ `.omk/memory/graph-state.json` — critical; local ontology graph database
+- ✅ `.kimi/skills` — support; Kimi skill directory
+- ✅ `.agents/skills` — support; portable skill directory
+
+## Recovery Command
+```bash
+omk init
+omk doctor
+```