@kodevibe/harness 0.9.3 → 0.9.5

package/README.ko.md

@@ -93,7 +93,7 @@ kode:harness는 세 가지 메커니즘으로 해결합니다:
 | 🧭 **Navigation Dispatcher** | 5개 파이프라인을 따라 다음 단계 프롬프트를 자동 안내 |
 | 📝 **상태 영속성** | LLM 세션 간 프로젝트 지식을 유지하는 5개 마크다운 파일 |
 | 🔄 **5개 파이프라인** | 🟢 신규 → 🔵 계속 → 🔴 버그 수정 → 🟡 방향 전환 → 🟣 Crew 기반 |
-| 🛠️ **10개 스킬** | 단계별 절차: setup, debug, breakdown, review, pivot 등 |
+| 🛠️ **11개 스킬** | 단계별 절차: setup, debug, breakdown, review, pivot, state-check 등 |
 | 🤖 **4개 에이전트** | 역할 기반 페르소나: pm, reviewer, lead, architect |
 | ⚠️ **실패 패턴** | 세션 간 같은 실수를 방지하는 프로젝트별 실패 기록 |
 | 📋 **Decision Log** | 결정의 이유를 기록해 LLM이 확정된 사항을 재논의하지 않도록 방지 |
@@ -116,10 +116,10 @@ npx @kodevibe/harness validate  # state 파일에 실제 내용 확인
 |-----|---------------------|------|----------|
 | **VS Code Copilot** | `.github/copilot-instructions.md` | `.github/skills/*/SKILL.md` | `.github/agents/*.agent.md` |
 | **Claude Code** | `CLAUDE.md` (+ `.claude/rules/core.md`) | `.claude/skills/*/SKILL.md` | `.claude/agents/*.md` |
-| **Cursor** | `.cursor/rules/core.mdc` (+ `AGENTS.md`) | `.cursor/skills/*/SKILL.md` | `.cursor/agents/*.md` |
-| **Codex** | `AGENTS.md` | `.agents/skills/*/SKILL.md` | `.codex/agents/*.md` |
+| **Cursor** | `.cursor/rules/core.mdc` (+ `AGENTS.md`) | `.agents/skills/*/SKILL.md` (cross-tool) | `.cursor/rules/<agent>.mdc` |
+| **Codex** | `AGENTS.md` | `.agents/skills/*/SKILL.md` | `.codex/agents/*.toml` |
 | **Windsurf** | `.windsurf/rules/core.md` | `.windsurf/skills/*/SKILL.md` | *(스킬로 설치)* |
-| **Antigravity** | `GEMINI.md` (+ `AGENTS.md`) | `.gemini/skills/*/SKILL.md` | `.gemini/agents/*.md` |
+| **Antigravity** | `AGENTS.md` | `.agents/skills/*/SKILL.md` (cross-tool) | `.agents/rules/<agent>.md` |
 
 모든 IDE에 `docs/` 디렉토리에 State 파일(`project-state.md`, `project-brief.md`, `features.md`, `failure-patterns.md`, `dependency-map.md`)도 함께 설치됩니다.
 
@@ -244,7 +244,7 @@ npx @kodevibe/harness init --team
 
 ## Iron Laws
 
-8개 규칙이 모든 스킬과 에이전트에 적용됩니다. kode:harness로 관리되는 프로젝트의 품질 근간을 형성합니다.
+10개 규칙이 모든 스킬과 에이전트에 적용됩니다. kode:harness로 관리되는 프로젝트의 품질 근간을 형성합니다.
 
 | # | 규칙 | 적용 대상 |
 |---|------|----------|
@@ -292,7 +292,7 @@ Bootstrap이 `docs/crew/`, `docs/PM/`, `docs/Analyst/`, `docs/ARB/`에서 crew
 | 의존성 | Node 20+ | Bun + Node + Playwright | Node 18+ | **Zero** |
 | IDE 지원 | 20+ (installer) | 5 (setup --host) | 13 (runtime select) | 6 (네이티브 포맷) |
 | 방향 관리 | ❌ | ❌ | ❌ | ✅ (Direction Guard + pivot + Decision Log) |
-| Iron Laws (코드 품질 규칙) | ❌ | ❌ | ❌ | ✅ (8개 규칙이 스킬에 임베딩) |
+| Iron Laws (코드 품질 규칙) | ❌ | ❌ | ❌ | ✅ (10개 규칙이 스킬에 임베딩) |
 | Cold start | ❌ | ❌ | `/gsd-new-project` | ✅ (`setup` 스킬) |
 | 태스크당 컨텍스트 | 4-6 파일 | 1 파일 | 매번 200k 플랜 | **2-3 파일 (136줄 디스패처)** |
 
@@ -300,7 +300,7 @@ Bootstrap이 `docs/crew/`, `docs/PM/`, `docs/Analyst/`, `docs/ARB/`에서 crew
 
 ## 로드맵
 
-kode:harness는 현재 **v0.9.2** — state-check 스킬, Iron Law #10 (Self-Verify), Confirmation Gate Defaults, 멀티 IDE 설치 수정, Crew 모드용 CI Artifact Index 도입.
+kode:harness는 현재 **v0.9.5** — 경량성 예산 재교정(40K/1500/2500), reviewer.md의 Iron Laws 정합성 수정, `harness/core-rules.md` ↔ `.github/copilot-instructions.md` 디스패처 동기화. v0.9.4는 6개 IDE 어댑터 공식 문서 정합(Antigravity `.agents/`, Codex `.toml`, Cursor `.cursor/rules/`).
 
 | 단계 | 버전 | 상태 | 초점 |
 |------|------|------|------|
@@ -309,7 +309,9 @@ kode:harness는 현재 **v0.9.2** — state-check 스킬, Iron Law #10 (Self-Ver
 | **Flexibility** | v0.7.x | ✅ 완료 | 팀 컨벤션을 project-brief.md에 위임, prescriptive 규칙 제거 |
 | **Navigation** | v0.8.x | ✅ 완료 | 🧭 Navigation Dispatcher, 5개 파이프라인, Crew Artifact Integration, 100점 품질 감사, Confirm-First 게이트, Wave-Level Pacing, Recalculating Mode |
 | **Naming** | v0.9.0 | ✅ 완료 | 스킬/에이전트 네이밍 재설계 — 직관성과 발견성 강화 |
-| **Self-Verify** | v0.9.2 | ✅ 현재 | state-check 스킬, Iron Law #10, Confirmation Gate Defaults, 멀티 IDE 수정, CI Artifact Index |
+| **Self-Verify** | v0.9.2 | ✅ 완료 | state-check 스킬, Iron Law #10, Confirmation Gate Defaults, 멀티 IDE 수정, CI Artifact Index |
+| **IDE Realignment** | v0.9.4 | ✅ 완료 | 6개 IDE 어댑터 공식 문서 정합; Antigravity `.agents/`, Codex `.toml`, Cursor `.cursor/rules/`; release 스킬 Step 6.5 + qa-check.sh §10 회귀 가드 |
+| **Consistency & Budget** | v0.9.5 | ✅ 현재 | reviewer.md Iron Laws stale 수정, 디스패처 동기화, 경량성 예산 재교정(40K/1500/2500) 및 근거 기록 |
 | **Validation** | v1.0 | 🔜 다음 | 실사용 검증, 사용자 피드백 수집 |
 
 ### 다음 단계

package/README.md

@@ -93,7 +93,7 @@ kode:harness solves this with three mechanisms:
 | 🧭 **Navigation Dispatcher** | Turn-by-Turn navigation through 5 pipelines with copy-paste next-step prompts |
 | 📝 **State Persistence** | 5 markdown files that persist project knowledge across LLM sessions |
 | 🔄 **5 Pipelines** | 🟢 New Dev → 🔵 Continue → 🔴 Bug Fix → 🟡 Direction Change → 🟣 Crew-Driven |
-| 🛠️ **10 Skills** | Step-by-step procedures: setup, debug, breakdown, review, pivot, and more |
+| 🛠️ **11 Skills** | Step-by-step procedures: setup, debug, breakdown, review, pivot, state-check, and more |
 | 🤖 **4 Agents** | Role-based personas: pm, reviewer, lead, architect |
 | ⚠️ **Failure Patterns** | Project-specific failure log that prevents repeat mistakes across sessions |
 | 📋 **Decision Log** | Records why decisions were made so LLMs don't re-debate settled choices |
@@ -114,10 +114,10 @@ npx @kodevibe/harness validate  # verify state files have real content
 |-----|----------------------|--------|--------|
 | **VS Code Copilot** | `.github/copilot-instructions.md` | `.github/skills/*/SKILL.md` | `.github/agents/*.agent.md` |
 | **Claude Code** | `CLAUDE.md` (+ `.claude/rules/core.md`) | `.claude/skills/*/SKILL.md` | `.claude/agents/*.md` |
-| **Cursor** | `.cursor/rules/core.mdc` (+ `AGENTS.md`) | `.cursor/skills/*/SKILL.md` | `.cursor/agents/*.md` |
-| **Codex** | `AGENTS.md` | `.agents/skills/*/SKILL.md` | `.codex/agents/*.md` |
+| **Cursor** | `.cursor/rules/core.mdc` (+ `AGENTS.md`) | `.agents/skills/*/SKILL.md` (cross-tool) | `.cursor/rules/<agent>.mdc` |
+| **Codex** | `AGENTS.md` | `.agents/skills/*/SKILL.md` | `.codex/agents/*.toml` |
 | **Windsurf** | `.windsurf/rules/core.md` | `.windsurf/skills/*/SKILL.md` | *(agents installed as skills)* |
-| **Antigravity** | `GEMINI.md` (+ `AGENTS.md`) | `.gemini/skills/*/SKILL.md` | `.gemini/agents/*.md` |
+| **Antigravity** | `AGENTS.md` | `.agents/skills/*/SKILL.md` (cross-tool) | `.agents/rules/<agent>.md` |
 
 All IDEs also get state files (`project-state.md`, `project-brief.md`, `features.md`, `failure-patterns.md`, `dependency-map.md`) in the `docs/` directory.
 
@@ -218,7 +218,7 @@ npx @kodevibe/harness init --team
 
 ## Iron Laws
 
-These 8 rules are enforced across all skills and agents. They form the quality backbone of every kode:harness project managed with harness engineering.
+These 10 rules are enforced across all skills and agents. They form the quality backbone of every kode:harness project managed with harness engineering.
 
 | # | Law | Enforced By |
 |---|-----|-------------|
@@ -268,13 +268,13 @@ Original crew documents are **never modified**. Only the index and tracker are c
 | Dependencies | Node 20+ | Bun + Node + Playwright | Node 18+ | Zero |
 | IDE support | 20+ (installer) | 5 (setup --host) | 13 (runtime select) | 6 (native format) |
 | Direction management | ❌ | ❌ | ❌ | ✅ (Direction Guard + pivot + Decision Log) |
-| Iron Laws (code quality rules) | ❌ | ❌ | ❌ | ✅ (8 laws embedded in skills) |
+| Iron Laws (code quality rules) | ❌ | ❌ | ❌ | ✅ (10 laws embedded in skills) |
 | Cold start | ❌ | ❌ | `/gsd-new-project` | ✅ (`setup` skill) |
 | Context per task | 4-6 files | 1 file | Fresh 200k per plan | 2-3 files (136-line dispatcher) |
 
 ## Roadmap
 
-kode:harness is at **v0.9.2** — state-check skill, Iron Law #10 (Self-Verify), Confirmation Gate Defaults, multi-IDE installation fixes, and CI Artifact Index for crew mode.
+kode:harness is at **v0.9.5** — lightness budget recalibrated (40K/1500/2500), Iron Laws consistency fix in reviewer.md, dispatcher synchronization between `harness/core-rules.md` and `.github/copilot-instructions.md`. v0.9.4 brought all 6 IDE adapters into alignment with official documentation (Antigravity `.agents/`, Codex `.toml`, Cursor `.cursor/rules/`).
 
 | Phase | Version | Status | Focus |
 |---|---|---|---|
@@ -283,7 +283,9 @@ kode:harness is at **v0.9.2** — state-check skill, Iron Law #10 (Self-Verify),
 | **Flexibility** | v0.7.x | ✅ Done | Delegate team conventions to project-brief.md, remove prescriptive rules |
 | **Navigation** | v0.8.x | ✅ Done | 🧭 Navigation Dispatcher, 5 Pipelines, Crew Artifact Integration, 100-point quality audit, Confirm-First gate, Wave-Level Pacing, Recalculating Mode |
 | **Naming** | v0.9.0 | ✅ Done | Skill/agent naming redesign for clarity and discoverability |
-| **Self-Verify** | v0.9.2 | ✅ Current | state-check skill, Iron Law #10, Confirmation Gate Defaults, multi-IDE fix, CI Artifact Index |
+| **Self-Verify** | v0.9.2 | ✅ Done | state-check skill, Iron Law #10, Confirmation Gate Defaults, multi-IDE fix, CI Artifact Index |
+| **IDE Realignment** | v0.9.4 | ✅ Done | All 6 IDE adapters aligned with official docs; Antigravity `.agents/`, Codex `.toml`, Cursor `.cursor/rules/`; release skill Step 6.5 + qa-check.sh §10 regression guards |
+| **Consistency & Budget** | v0.9.5 | ✅ Current | Iron Laws stale-copy fix (reviewer.md), dispatcher sync (core-rules.md ↔ copilot-instructions.md), lightness budgets recalibrated (40K/1500/2500) with rationale |
 | **Validation** | v1.0 | 🔜 Next | Real-world project adoption, user feedback collection |
 
 ### What's Next

package/harness/agents/pm.md

@@ -61,8 +61,6 @@ Read `docs/agent-memory/pm.md` for past learnings:
 
 Apply these insights when creating the implementation plan. If the memory file is empty or contains only placeholders, skip this step.
 
-> **Team Mode**: In Team mode, agent memory is personal (`.harness/agent-memory/`). Each developer accumulates their own planning insights.
-
 ### Step 0.7: Feature Roadmap Planning (Draft & Correct)
 
 **Trigger**: `docs/project-brief.md`에 `## Feature Roadmap` 섹션이 없을 때
@@ -87,12 +85,9 @@ Apply these insights when creating the implementation plan. If the memory file i
    - [ ] F-004: ...
    ```
 3. 사용자에게 초안을 제시한다: **"이 Feature Roadmap을 검토하고, 추가/삭제/순서 변경을 알려주세요."**
-   > **물어보지 말고, 초안을 만들어서 교정받는다.** 빈 칸을 채우는 것보다 틀린 것을 고치는 것이 쉽다.
 4. 사용자 교정을 반영한 최종 Roadmap을 `docs/project-brief.md`에 `## Feature Roadmap` 섹션으로 기록한다
 5. Feature Roadmap이 확정되면 아래 "For New Feature" 절차로 진행한다
 
-> **pivot 이후**: pivot이 `project-brief.md`를 업데이트하면, pm은 변경된 Roadmap을 읽고 Checkpoint에서 반영한다.
-
 ### For New Feature
 
 1. Read `docs/project-brief.md` to understand project vision, goals, **non-goals**, and **Decision Log**
@@ -133,12 +128,11 @@ Apply these insights when creating the implementation plan. If the memory file i
    If no Crew Artifact Index → proceed with normal user-driven planning below.
 <!-- CREW_MODE_END -->
 
-3. **Direction Alignment**: Verify the requested feature against three checkpoints.
-   > This check intentionally duplicates architect’s direction validation (Step 2). The redundancy is by design: architect validates STRUCTURAL proposals (module boundaries, layer rules), while pm validates FEATURE-level alignment (goals, non-goals, decisions). When both are used in the same session, this provides defense-in-depth.
-   - **Goal Alignment**: Does it serve a listed Goal? If no clear link → **warn but proceed**. Include the warning in the plan output under a `### Direction Alignment` section: `⚠️ Goal Alignment: [feature] does not directly map to listed goals`.
-   - **Non-Goal Violation**: Does it fall into Non-Goals? If yes → **stop and ask the user**. Do not proceed until the user confirms this is intentional (may need `pivot` skill).
-   - **Decision Consistency**: Does it contradict any Decision Log entry? If yes → **stop and warn**. Recommend running the `pivot` skill before proceeding.
-   If the request represents a clear direction change → **stop and require the `pivot` skill** before proceeding with any planning. Do not proceed even if the user insists — direction changes must be formally tracked.
+3. **Direction Alignment**: Verify against three checkpoints (architect validates STRUCTURE; pm validates FEATURE-level alignment):
+   - **Goal Alignment**: Serves a listed Goal? If no clear link → **warn but proceed**. Add `⚠️ Goal Alignment: [feature] does not directly map to listed goals` under `### Direction Alignment` in the plan output.
+   - **Non-Goal Violation**: Falls into Non-Goals? → **stop and ask the user**. May need `pivot`.
+   - **Decision Consistency**: Contradicts a Decision Log entry? → **stop and warn**. Recommend `pivot`.
+   If the request represents a clear direction change → **stop and require `pivot`** before planning. Do not proceed even if the user insists.
 3. Read `docs/features.md` to understand what already exists
 4. Read `docs/dependency-map.md` to understand current architecture
 5. Read `docs/project-state.md` for current Sprint context
@@ -153,7 +147,7 @@ Apply these insights when creating the implementation plan. If the memory file i
 13. **After user approves** → Update `docs/project-state.md` with the new Story
 14. **After user approves** → Update `docs/features.md` with the new feature entry
 
-> **State File Write Deferral**: Steps 13-14 execute ONLY after user confirms the plan. If the user rejects or requests changes, no state files are modified — the plan is revised and re-presented. This prevents state file pollution from rejected plans.
+State file writes (Steps 13-14) execute ONLY after user approval. Rejected plans never touch state.
 
 ### For Architecture Query
 
@@ -187,8 +181,6 @@ After producing ANY plan (New Feature, Refactor, or Crew-Driven), **do NOT proce
 4. **Only after approval** → execute **MANDATORY State File Write** (below), then output 🧭 Next Step pointing to `lead`
 5. If the user requests changes → revise the plan and re-confirm. **No state files are written until approval.**
 
-> **Why**: The pm is planning a route, not driving. The user must confirm the route before the engine starts. This prevents irreversible code changes based on a misunderstood plan.
-
 ### ⚠️ MANDATORY: Post-Approval State File Write
 
 **This section executes IMMEDIATELY after user approval. Do NOT skip. Do NOT output the 🧭 Next Step block until ALL writes below are complete.**
@@ -234,8 +226,6 @@ After the Post-Approval state writes complete, run the `state-check` skill:
 3. If state-check returns **WARN** → include the warnings in the plan output, then proceed
 4. If state-check returns **FAIL** → do NOT output STATUS: DONE. Fix the listed drift, then re-run state-check.
 
-> Iron Law #10 (Self-Verify) applies to every agent. The pm runs state-check **after** state writes — not before — because the writes are what create the consistency to verify.
-
 ## Output Format
 
 ### New Feature Plan
@@ -291,9 +281,8 @@ Use this format when Crew Artifact Index exists in project-brief.md. If no Artif
 
 ## Sprint Completion Checkpoint
 
-**Trigger**: 현재 Sprint의 모든 Story가 ✅ 완료되었을 때 (reviewer pass → pm checkpoint)
-
-이 절차는 Sprint 종료 시 자동으로 실행된다. "계속할까요?"가 아니라 **구체적 선택을 강제**한다.
+**Trigger**: 현재 Sprint의 모든 Story가 ✅ 완료되었을 때 (reviewer pass → pm checkpoint).
+Sprint 종료 시 자동 실행 — **구체적 선택을 강제**한다 ("계속할까요?" 가 아님).
 
 ### 절차
 
@@ -324,7 +313,7 @@ Use this format when Crew Artifact Index exists in project-brief.md. If no Artif
 
    🏁 마무리: 현재 상태로 프로젝트를 완료하고 wrap-up 진행
    ```
-   > **"계속"은 선택지에 없다.** 사용자는 구체적 기능을 골라야 하거나, 명시적으로 "마무리"를 선택해야 한다.
+   **"계속"은 선택지에 없다** — 사용자는 구체적 기능 또는 "마무리"를 명시적으로 선택해야 한다.
 
 4. **사용자 선택에 따라 분기**:
    - **기능 선택** → 선택된 기능으로 Sprint 계획 (위의 "For New Feature" 절차 진행)

package/harness/agents/reviewer.md

@@ -217,17 +217,7 @@ STATUS: DONE / DONE_WITH_CONCERNS / BLOCKED
 
 ## Embedded Rules
 
-These rules are enforced during every review:
-
-### Iron Laws
-1. **Mock Sync**: Interface change → mock updated in same commit (FP-001)
-2. **Type Check**: Verify constructor/factory parameters from source, not memory (FP-002)
-3. **Scope Compliance**: Changes must be within current Story scope (docs/project-state.md)
-4. **Security**: No credentials, passwords, or API keys in code or commits
-5. **3-Failure Stop**: Same approach failed 3 times → stop and report
-6. **Dependency Map**: New/modified module → docs/dependency-map.md updated
-7. **Feature Registry**: New feature → docs/features.md updated
-8. **Session Handoff**: Session end → docs/project-state.md Quick Summary updated
+These rules are enforced during every review. The full Iron Laws (10) are defined in `harness/core-rules.md` — reviewer enforces all of them. Below are review-specific rules that supplement the Iron Laws.
 
 ### Testing Rules
 - New feature = New test. No feature code without tests.

package/harness/skills/release.md

@@ -83,6 +83,25 @@ This step references the indexed company guide; it does not embed the guide cont
 2. If yes → verify they include entries for all changes since last release
 3. If no → generate a summary from `git log --oneline <last-tag>..HEAD`
 
+### Step 6.5: IDE Adapter 공식 문서 정합성 (IDE-targeting 프로젝트만)
+
+If the project ships IDE-specific adapters or generators (e.g., `src/init.js` with multiple `generate<IDE>` functions, OR distributes templates installed under IDE-specific directories like `.cursor/`, `.codex/`, `.windsurf/`, `.agents/`):
+
+1. **List all IDEs supported** by the project (read source — do not guess).
+2. For **each IDE**, fetch the **official current documentation** for that IDE's agent/rules/skills layout. Do not rely on training data.
+3. **Diff** the project's generated paths/file formats against the official docs:
+   - File extensions (e.g., Codex requires `.toml`, not `.md`)
+   - Directory locations (e.g., Antigravity uses `.agents/`, not `.gemini/`)
+   - Frontmatter schema (required fields per IDE)
+   - Cross-tool standards (e.g., `AGENTS.md`, `.agents/skills/`)
+4. If **any drift** is detected → mark **NOT_READY**. Block the release until adapters match official docs.
+5. Verify regression tests assert the **absence** of stale paths (e.g., `assert(!exists('.gemini/'))` if the IDE moved off Gemini).
+6. Cite the official doc URL inline in the generator code as a comment (source-of-truth pointer).
+
+**Rationale**: IDE vendors change their layouts (e.g., Cursor's skills location, Antigravity's `.gemini/` → `.agents/` migration). Without this gate, every IDE update silently breaks `harness init` for that IDE's users.
+
+If the project does not ship IDE adapters → skip this step entirely.
+
 ## Output Format
 
 ```

package/harness/skills/setup.md

@@ -278,28 +278,6 @@ For projects with fewer than 3 modules (e.g., single-file scripts, small CLI too
 - `breakdown` Waves may collapse into a single Wave — skip Wave-level pacing
 - Consider a simplified workflow: `setup → pm → [code] → reviewer → wrap-up` (skip lead for single-story projects)
 
-## Embedded Knowledge
-
-### Session Bootstrap Protocol
-When starting a NEW session (not during setup), read these files in order:
-1. `docs/project-state.md` — Quick Summary tells you where we left off
-2. `docs/features.md` — What features exist
-3. `docs/failure-patterns.md` — What mistakes to avoid
-4. `docs/project-brief.md` — Project vision and non-goals
-
-### Workflow Pipeline
-- New feature: `pm → [code] → reviewer → lead → wrap-up`
-- Bug fix: `debug → [fix] → sync-tests → reviewer → wrap-up`
-- Session lifecycle: `lead ("where are we?") → [work] → wrap-up`
-
-### State File Size Limits
-- docs/project-brief.md: Max 200 lines
-- docs/project-state.md: Max 300 lines
-- docs/failure-patterns.md: Max 50 patterns
-- docs/dependency-map.md: Max 100 modules
-- docs/features.md: Max 50 features
-- docs/agent-memory/*.md: Max 100 lines each
-
 ## Anti-patterns
 
 | Anti-pattern | Correct Approach |

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@kodevibe/harness",
-  "version": "0.9.3",
+  "version": "0.9.5",
   "description": "kode:harness — harness engineering for keeping every developer's AI aligned on one project direction.",
   "keywords": [
     "llm",

package/src/init.js

@@ -78,13 +78,16 @@ function hasFrameworkMarker(content) {
 }
 
 function hasIdeLayout(targetDir, ide) {
+  // IDE-unique layout markers — used to disambiguate when multiple IDEs share
+  // common files (e.g., AGENTS.md, .agents/skills/). Each marker MUST be unique
+  // to its IDE and aligned with the IDE's official docs.
   const requiredByIde = {
-    vscode: '.github/skills/setup/SKILL.md',
-    claude: '.claude/skills/setup/SKILL.md',
-    cursor: '.cursor/skills/setup/SKILL.md',
-    codex: '.agents/skills/setup/SKILL.md',
-    windsurf: '.windsurf/skills/setup/SKILL.md',
-    antigravity: '.gemini/skills/setup/SKILL.md',
+    vscode: '.github/skills/setup/SKILL.md',          // VS Code: .github/skills/ (official)
+    claude: '.claude/skills/setup/SKILL.md',          // Claude Code: .claude/skills/ (official)
+    cursor: '.cursor/rules/core.mdc',                 // Cursor: .cursor/rules/*.mdc (official)
+    codex: '.codex/agents/reviewer.toml',             // Codex CLI: .codex/agents/*.toml (official)
+    windsurf: '.windsurf/skills/setup/SKILL.md',      // Windsurf: .windsurf/skills/ (official)
+    antigravity: '.agents/rules/core.md',             // Antigravity: .agents/rules/ (official)
   };
 
   const requiredPath = requiredByIde[ide];
@@ -271,37 +274,55 @@ function generateClaude(targetDir, overwrite, mode = 'solo', crew = false) {
 }
 
 function generateCursor(targetDir, overwrite, mode = 'solo', crew = false) {
-  // .cursor/rules/core.mdc — dispatcher only (always active)
+  // Cursor official docs: https://cursor.com/docs/context/rules
+  // Officially supported surfaces ONLY:
+  //   - .cursor/rules/*.mdc (with frontmatter: alwaysApply, description, globs)
+  //   - AGENTS.md at project root (and nested directories)
+  // Cursor does NOT publicly document custom skills/agents directories, so we
+  // emit Skills via the open Agent Skills standard path .agents/skills/
+  // (also recognized by VS Code, Codex, Antigravity), and skip a Cursor-specific
+  // agents directory entirely.
   const coreRules = resolveContent(readTemplate('core-rules.md'), mode, crew);
   const coreMdc =
     '---\ndescription: kode:harness dispatcher — workflow guidance and state file references\nalwaysApply: true\n---\n\n' +
     coreRules;
   writeFile(targetDir, '.cursor/rules/core.mdc', coreMdc, true);
 
-  // AGENTS.md — Cursor CLI also reads project-root AGENTS.md as a rule
+  // AGENTS.md — Cursor reads project-root AGENTS.md as agent instructions
   writeFile(targetDir, 'AGENTS.md', coreRules, true);
 
-  // Skills (.cursor/skills — invokable by mentioning skill name)
-  writeSkills(targetDir, '.cursor/skills', true, mode, crew);
+  // Skills (.agents/skills — open Agent Skills standard, cross-tool)
+  writeSkills(targetDir, '.agents/skills', true, mode, crew);
 
-  // Agents (.cursor/agents/ — Cursor subagent definition files)
-  writeAgentsAsMd(targetDir, '.cursor/agents', true, mode, crew);
+  // Agents as additional rules (.cursor/rules/) so they auto-load with Cursor
+  // and are user-invocable via @rule-name. We use .md plain (no frontmatter
+  // beyond alwaysApply omitted → default "manual" activation).
+  for (const agent of AGENTS) {
+    const content = resolveContent(readTemplate(agent.file), mode, crew);
+    const ruleMd =
+      `---\ndescription: ${agent.desc}\nalwaysApply: false\n---\n\n` +
+      content;
+    writeFile(targetDir, `.cursor/rules/${agent.id}.mdc`, ruleMd, true);
+  }
 
   // State files (respect user's --overwrite for data files)
   writeStateFiles(targetDir, overwrite, mode, crew);
 }
 
 function generateCodex(targetDir, overwrite, mode = 'solo', crew = false) {
-  // AGENTS.md — Codex CLI's canonical project instructions file (the only file
-  // Codex CLI auto-loads). All skill/agent references must be discoverable from here.
+  // Codex CLI official docs:
+  //   - AGENTS.md: https://developers.openai.com/codex/guides/agents-md
+  //   - Skills: https://developers.openai.com/codex/skills (.agents/skills/SKILL.md)
+  //   - Subagents: https://developers.openai.com/codex/subagents (.codex/agents/*.toml)
+  // AGENTS.md — Codex CLI's canonical project instructions file (auto-loaded).
   writeFile(targetDir, 'AGENTS.md', resolveContent(readTemplate('core-rules.md'), mode, crew), true);
 
-  // Skills (SKILL.md with frontmatter — invokable via $skill-name)
+  // Skills (.agents/skills/<name>/SKILL.md — official cross-tool standard)
   writeSkills(targetDir, '.agents/skills', true, mode, crew);
 
-  // Agents (.codex/agents/ — markdown subagent files with YAML frontmatter,
-  // matching the Cursor/Claude convention used as a Codex compat directory)
-  writeAgentsAsMd(targetDir, '.codex/agents', true, mode, crew);
+  // Subagents (.codex/agents/*.toml — Codex CLI's official format with
+  // name, description, developer_instructions fields)
+  writeAgentsAsToml(targetDir, '.codex/agents', true, mode, crew);
 
   // State files (respect user's --overwrite for data files)
   writeStateFiles(targetDir, overwrite, mode, crew);
@@ -326,20 +347,31 @@ function generateWindsurf(targetDir, overwrite, mode = 'solo', crew = false) {
 }
 
 function generateAntigravity(targetDir, overwrite, mode = 'solo', crew = false) {
+  // Antigravity official docs:
+  //   - Skills: https://antigravity.google/docs/skills (.agents/skills/<name>/SKILL.md)
+  //   - Rules: https://antigravity.google/docs/rules-workflows (.agents/rules/*.md)
+  // Note: Antigravity does NOT recognize a project-root GEMINI.md — the global
+  // GEMINI.md lives only at ~/.gemini/GEMINI.md. We emit AGENTS.md for cross-tool
+  // compat (harmless to Antigravity, useful when Codex/Cursor share the repo).
   const coreRules = resolveContent(readTemplate('core-rules.md'), mode, crew);
 
-  // GEMINI.md — Gemini CLI / Antigravity's canonical project context file
-  writeFile(targetDir, 'GEMINI.md', coreRules, true);
-
-  // AGENTS.md — Antigravity also follows the AGENTS.md convention shared by
-  // Codex / Cursor CLI; emitting it broadens compatibility with no downside
+  // AGENTS.md — cross-tool agent instructions (recognized by Codex/Cursor;
+  // ignored by Antigravity but provides compatibility for mixed teams)
   writeFile(targetDir, 'AGENTS.md', coreRules, true);
 
-  // Skills (.gemini/skills/ — SKILL.md format)
-  writeSkills(targetDir, '.gemini/skills', true, mode, crew);
+  // .agents/rules/core.md — workspace rules (Antigravity's official location)
+  writeFile(targetDir, '.agents/rules/core.md', coreRules, true);
 
-  // Agents (.gemini/agents/ — Gemini CLI subagent definition files)
-  writeAgentsAsMd(targetDir, '.gemini/agents', true, mode, crew);
+  // Skills (.agents/skills/<name>/SKILL.md — Antigravity's official location)
+  writeSkills(targetDir, '.agents/skills', true, mode, crew);
+
+  // Agents → Rules (Antigravity has no separate agents directory; agents
+  // are persistent rule prompts. We emit one rule file per agent under
+  // .agents/rules/, with alwaysApply implied — manual @-mention also works.)
+  for (const agent of AGENTS) {
+    const content = resolveContent(readTemplate(agent.file), mode, crew);
+    writeFile(targetDir, `.agents/rules/${agent.id}.md`, content, true);
+  }
 
   // State files (respect user's --overwrite for data files)
   writeStateFiles(targetDir, overwrite, mode, crew);
@@ -430,7 +462,7 @@ function detectExistingInstall(targetDir) {
     ['.claude/rules/core.md', 'claude'],
     ['.cursor/rules/core.mdc', 'cursor'],
     ['.windsurf/rules/core.md', 'windsurf'],
-    ['GEMINI.md', 'antigravity'],
+    ['.agents/rules/core.md', 'antigravity'],
   ];
   // Only count as existing if the file contains a framework marker (not from other frameworks)
   const existingIde = ideMarkers.filter(([f, ide]) => {
@@ -546,7 +578,8 @@ function runDoctor(targetDir) {
     ['.claude/rules/core.md', 'claude'],
     ['.cursor/rules/core.mdc', 'cursor'],
     ['.windsurf/rules/core.md', 'windsurf'],
-    ['GEMINI.md', 'antigravity'],
+    ['.agents/rules/core.md', 'antigravity'],
+    ['.codex/agents/reviewer.toml', 'codex'],
     ['AGENTS.md', 'codex'],
   ];