@wooojin/forgen 0.2.0 → 0.3.0

package/commands/ship.md

@@ -0,0 +1,277 @@
+---
+name: ship
+description: This skill should be used when the user asks to "ship, 배포, 릴리스, release". 비대화형 자동 릴리스 파이프라인 — 테스트, 리뷰, 버전, CHANGELOG, PR을 원커맨드로.
+argument-hint: "[patch|minor|major]"
+model: inherit
+disable-model-invocation: true
+allowed-tools:
+  - Read
+  - Grep
+  - Glob
+  - Bash
+  - Agent
+  - Edit
+  - Write
+triggers:
+  - "ship"
+  - "배포"
+  - "릴리스"
+  - "release"
+  - "릴리즈"
+  - "배포해줘"
+---
+
+<Purpose>
+"사용자가 /ship이라고 했으면 실행하라."
+
+테스트 -> 리뷰 -> 버전 범프 -> CHANGELOG -> 커밋 -> PR 생성까지의 릴리스 파이프라인을 완전 자동화합니다.
+사용자에게 묻지 않습니다. 중단 사유가 아닌 한 끝까지 진행합니다.
+
+### 중단 사유 (이것만 멈춘다)
+- main/master 브랜치에서 직접 실행 (abort)
+- 해결 불가능한 머지 충돌
+- 테스트 실패
+- CRITICAL 리뷰 발견
+- MAJOR 버전 범프 결정 (사용자 확인 필요)
+
+이 외의 모든 상황은 자동 판단하여 진행합니다.
+</Purpose>
+
+<Compound_Integration>
+## Compound-In: 이전 릴리스 이슈 로드
+
+Step 1.5에서 실행합니다.
+
+```
+compound-search("ship release 이슈 배포 실패")
+```
+
+이전 릴리스에서 발생한 문제 패턴을 확인합니다.
+결과가 있으면 해당 체크포인트를 강화합니다.
+
+## Compound-Out: 릴리스 이슈 기록
+
+Step 10에서 실행합니다.
+릴리스 중 발생한 이슈가 있었으면 compound troubleshoot로 기록을 제안합니다.
+</Compound_Integration>
+
+<Steps>
+## Step 0: 플랫폼 감지 + 베이스 브랜치 감지
+
+```bash
+# 플랫폼 감지
+if command -v gh &>/dev/null && gh repo view &>/dev/null; then
+  PLATFORM="github"
+elif command -v glab &>/dev/null; then
+  PLATFORM="gitlab"
+else
+  PLATFORM="unknown"
+fi
+
+# 베이스 브랜치 감지
+BASE=$(git symbolic-ref refs/remotes/origin/HEAD 2>/dev/null | sed 's@^refs/remotes/origin/@@')
+if [ -z "$BASE" ]; then
+  BASE=$(git branch -r | grep -E 'origin/(main|master)' | head -1 | sed 's@.*origin/@@' | tr -d ' ')
+fi
+```
+
+## Step 1: Pre-flight 체크
+
+```bash
+BRANCH=$(git branch --show-current)
+
+# main/master -> ABORT
+if [ "$BRANCH" = "main" ] || [ "$BRANCH" = "master" ]; then
+  echo "ABORT: main/master 브랜치에서 직접 릴리스할 수 없습니다."
+  exit 1
+fi
+
+git status --porcelain
+git diff --stat
+git diff --staged --stat
+```
+
+- main/master -> ABORT (새 브랜치 제안하지 않음. 그냥 중단.)
+- 미커밋 변경 -> 자동 커밋 ("chore: stage uncommitted changes before release")
+
+## Step 1.5: Compound-In
+
+```
+compound-search("ship release 배포 이슈")
+```
+
+이전 문제 패턴이 있으면 체크포인트 강화. 없으면 조용히 진행.
+
+## Step 2: 베이스 브랜치 머지
+
+```bash
+git fetch origin $BASE
+git merge origin/$BASE --no-edit
+```
+
+- 충돌 -> ABORT ("머지 충돌 발생. 수동 해결 후 다시 /ship하세요.")
+
+## Step 3: 테스트 실행
+
+```bash
+# 자동 감지: package.json -> vitest/jest -> pytest -> cargo test -> go test
+if [ -f package.json ]; then
+  if grep -q '"test"' package.json; then npm test
+  elif command -v vitest &>/dev/null; then npx vitest run
+  fi
+elif [ -f pytest.ini ] || [ -f pyproject.toml ]; then pytest
+elif [ -f Cargo.toml ]; then cargo test
+elif [ -f go.mod ]; then go test ./...
+fi
+```
+
+- 전체 통과 -> Step 3.5
+- 1개라도 실패 -> ABORT
+
+## Step 3.5: 테스트 커버리지 감사 (선택)
+
+```bash
+npx vitest run --coverage 2>/dev/null || true
+```
+
+커버리지는 정보 제공용. 미달로 중단하지 않음.
+
+## Step 4: Pre-landing 리뷰
+
+ch-code-reviewer 에이전트(READ-ONLY) 위임.
+
+```bash
+git diff $BASE...HEAD
+```
+
+- CRITICAL -> ABORT
+- AUTO-FIX (데드 코드, 미사용 import, stale 주석) -> 직접 수정 + 커밋
+- 판단 필요 -> 사용자에게 질문
+- APPROVED -> Step 5
+
+## Step 5: 버전 범프
+
+### 인수가 있으면: 인수 그대로 사용
+### 없으면 자동 판단:
+- < 50줄 -> PATCH (자동)
+- 50줄+ 기능 없음 -> PATCH (자동)
+- 기능 신호 (feat/, 새 라우트) -> MINOR (확인)
+- Breaking change -> MAJOR (확인)
+
+```bash
+npm version {patch|minor|major} --no-git-tag-version
+```
+
+## Step 6: CHANGELOG 생성
+
+커밋을 주제별 그룹핑 -> CHANGELOG.md 상단에 추가.
+
+## Step 7: 릴리스 커밋
+
+```bash
+git add package.json CHANGELOG.md
+git commit -m "release: v{version}"
+```
+
+## Step 8: Push
+
+```bash
+git push -u origin $(git branch --show-current)
+```
+
+force push 절대 안 함.
+
+## Step 9: PR 생성
+
+```bash
+gh pr create --title "release: v{version}" --body "..."
+```
+
+## Step 10: Compound-Out
+
+이슈가 있었으면 compound troubleshoot로 기록 제안.
+</Steps>
+
+<Review_Readiness_Dashboard>
+```
++============================================+
+|         REVIEW READINESS                    |
++============================================+
+| Check              | Status    | Required  |
+|--------------------|-----------|-----------|
+| Tests              | {N} PASS  | YES       |
+| Code Review        | {result}  | YES       |
+| Build              | {result}  | YES       |
+| Coverage           | {N}%      | no        |
+| Base Merge         | CLEAN     | YES       |
++============================================+
+| VERDICT: {READY TO SHIP / BLOCKED}         |
++============================================+
+```
+</Review_Readiness_Dashboard>
+
+<Verification_Gate>
+## 검증의 철칙 (IRON LAW)
+
+```
+"아마 될 거야"        -> 실행해라.
+"확신이 있다"         -> 확신은 증거가 아니다.
+"아까 테스트했는데"    -> 코드가 바뀌었다. 다시 테스트해라.
+"사소한 변경이라"      -> 사소한 변경이 프로덕션을 깨뜨린다.
+"이건 리뷰 안 해도"   -> 모든 diff는 리뷰한다.
+```
+
+실행 결과가 곧 증거입니다. 실행하지 않은 검증은 존재하지 않습니다.
+</Verification_Gate>
+
+<Failure_Modes>
+**테스트 실패 상태에서 진행**: 0 failures만 통과.
+**main 브랜치에서 직접 실행**: ABORT. 새 브랜치 제안도 안 함.
+**머지 충돌 자동 해결**: 충돌은 사용자가 수동 해결.
+**CRITICAL 리뷰 이슈 무시**: CRITICAL 1개라도 있으면 ABORT.
+**force push**: 어떤 상황에서도 사용하지 않음.
+**불필요한 질문**: /ship은 비대화형. 중단 사유만 보고.
+</Failure_Modes>
+
+<Output>
+## 성공 시
+
+```
+SHIP COMPLETE / 배포 완료
+=========================
+Version: {old} -> {new} ({type})
+Tests:   {N} passed, 0 failed
+Review:  APPROVED ({N} auto-fixed)
+PR:      #{number}
+URL:     {url}
+```
+
+## 실패 시
+
+```
+SHIP ABORTED / 배포 중단
+=========================
+Step:    {실패한 Step}
+Reason:  {중단 사유}
+Action:  {다음 행동}
+```
+</Output>
+
+<Policy>
+- 테스트 통과는 필수. 예외 없음.
+- CRITICAL 리뷰 이슈 -> 사용자 동의 없이 진행하지 않음.
+- main 브랜치 직접 릴리스 = ABORT.
+- 중단 사유가 아닌 한 묻지 않고 끝까지 진행.
+- 실행 결과 = 증거. 추측 != 검증.
+- force push 금지.
+- AUTO-FIX는 데드 코드/미사용 import/stale 주석에만 적용.
+</Policy>
+
+<Arguments>
+- `patch`: 버그 수정 릴리스 (0.0.X)
+- `minor`: 하위 호환 새 기능 릴리스 (0.X.0)
+- `major`: 하위 비호환 변경 릴리스 (X.0.0)
+- 생략 시: 변경 내용 자동 분석하여 결정 (major만 사용자 확인)
+</Arguments>
+
+$ARGUMENTS

package/dist/cli.js

@@ -9,11 +9,14 @@ import * as path from 'node:path';
 import { fileURLToPath } from 'node:url';
 import { prepareHarness, isFirstRun } from './core/harness.js';
 import { spawnClaudeWithResume } from './core/spawn.js';
+import { resolveLaunchContext } from './services/session.js';
 // global-config is used by harness internally
 const __dirname = path.dirname(fileURLToPath(import.meta.url));
 const pkgJson = JSON.parse(fs.readFileSync(path.resolve(__dirname, '..', 'package.json'), 'utf-8'));
 const PKG_VERSION = pkgJson.version ?? '0.0.0';
-const args = process.argv.slice(2);
+const launchContext = resolveLaunchContext(process.argv.slice(2));
+const args = launchContext.args;
+const runtime = launchContext.runtime;
 const commands = [
     {
         name: 'forge',
@@ -68,6 +71,14 @@ const commands = [
             }
         },
     },
+    {
+        name: 'dashboard',
+        description: 'Compound system dashboard with rich overview',
+        handler: async (_args) => {
+            const { handleDashboard } = await import('./core/dashboard.js');
+            await handleDashboard();
+        },
+    },
     {
         name: 'me',
         description: 'Personal dashboard (→ inspect profile)',
@@ -85,7 +96,7 @@ const commands = [
                 if (args.includes('--regenerate')) {
                     const { writeHooksJson } = await import('./hooks/hooks-generator.js');
                     const hooksDir = path.join(process.cwd(), 'hooks');
-                    const result = writeHooksJson(hooksDir, { cwd: process.cwd() });
+                    const result = writeHooksJson(hooksDir, { cwd: process.cwd(), runtime });
                     console.log(`[forgen] hooks.json regenerated: ${result.active} active, ${result.disabled} disabled`);
                 }
                 else {
@@ -219,14 +230,14 @@ async function main() {
 
   Setting up...`);
         }
-        let context = await prepareHarness(process.cwd());
+        let context = await prepareHarness(process.cwd(), { runtime });
         // 첫 실행 또는 프로필 없음 → 자동 온보딩 (interactive 환경)
         if (context.v1.needsOnboarding && process.stdin.isTTY) {
             console.log('\n  프로필이 없습니다. 온보딩을 시작합니다.\n');
             const { runOnboarding } = await import('./forge/onboarding-cli.js');
             await runOnboarding();
             // 온보딩 후 harness 재실행 (프로필 반영)
-            context = await prepareHarness(process.cwd());
+            context = await prepareHarness(process.cwd(), { runtime });
         }
         if (firstRun && !context.v1.needsOnboarding) {
             console.log(`
@@ -244,12 +255,16 @@ async function main() {
   ${dim}Code, forged for you.${reset}
   ${dim}Scope: v1(${context.v1.session?.quality_pack ?? 'onboarding needed'})${reset}
 `);
-        console.log('[forgen] Starting Claude Code...\n');
-        await spawnClaudeWithResume(args, context, () => prepareHarness(process.cwd()));
+        const runtimeLabel = runtime === 'codex' ? 'Codex' : 'Claude';
+        console.log(`[forgen] Starting ${runtimeLabel}...\n`);
+        await spawnClaudeWithResume(args, context, () => prepareHarness(process.cwd(), { runtime }), runtime);
     }
     catch (err) {
         const msg = err instanceof Error ? err.message : String(err);
-        if (msg.includes('ENOENT') && msg.includes('claude')) {
+        if (msg.includes('Codex is not installed')) {
+            console.error('[forgen] Codex is not installed.');
+        }
+        else if (msg.includes('Claude Code is not installed')) {
             console.error('[forgen] Claude Code not found. Install: npm install -g @anthropic-ai/claude-code');
         }
         else {
@@ -267,9 +282,10 @@ function printHelp() {
   The more you use Claude, the better it knows you.
 
   Usage:
-    forgen                          Start Claude Code (harness mode)
+    forgen                          Start runtime launcher (harness mode)
     forgen "prompt"                 Start with a prompt
     forgen --resume                 Resume previous session
+    forgen --runtime claude|codex   Select launch runtime
 
   Commands:
     forgen forge                    Personalize your coding profile
@@ -277,6 +293,7 @@ function printHelp() {
     forgen inspect [profile|rules|evidence|session]
                                     Inspect v1 state
     forgen compound                 Manage accumulated knowledge
+    forgen dashboard                Compound system dashboard
     forgen me                       Personal dashboard
     forgen init                     Initialize project
     forgen config hooks             Hook management
@@ -287,7 +304,7 @@ function printHelp() {
     forgen uninstall                Remove forgen
 
   Harness mode (default):
-    Wraps Claude Code with personalization, auto-compound, and safety hooks.
+    Wraps Claude (or Codex) with personalization, auto-compound, and safety hooks.
 `);
 }
 main().catch(() => {

package/dist/core/auto-compound-runner.js

@@ -483,6 +483,20 @@ ${sanitizedSummary.slice(0, 4000)}
     catch (e) {
         process.stderr.write(`[forgen-auto-compound] rule promotion: ${e instanceof Error ? e.message : String(e)}\n`);
     }
+    // Step 5: meta-learning (HyperAgents-inspired self-tuning)
+    try {
+        const { runMetaLearning } = await import('../engine/meta-learning/runner.js');
+        const metaResult = runMetaLearning(sessionId, cwd);
+        if (metaResult.qualityScore) {
+            process.stderr.write(`[forgen-meta] session quality: ${metaResult.qualityScore.overallScore}/100\n`);
+        }
+        if (metaResult.scopePromotions && metaResult.scopePromotions.length > 0) {
+            process.stderr.write(`[forgen-meta] promoted ${metaResult.scopePromotions.length} solution(s) to universal scope\n`);
+        }
+    }
+    catch (e) {
+        process.stderr.write(`[forgen-meta] ${e instanceof Error ? e.message : String(e)}\n`);
+    }
     // 완료 기록
     const statePath = path.join(FORGEN_HOME, 'state', 'last-auto-compound.json');
     fs.mkdirSync(path.dirname(statePath), { recursive: true });

package/dist/core/config-injector.d.ts

@@ -6,6 +6,7 @@
  *
  * Authoritative: docs/plans/2026-04-03-forgen-rule-renderer-spec.md
  */
+import type { RuntimeHost } from './types.js';
 /** 보안 규칙 (정적 — v1 GLOBAL_SAFETY_RULES와 동일 맥락) */
 export declare function generateSecurityRules(): string;
 /** 안티패턴 감지 규칙 (정적) */
@@ -36,7 +37,7 @@ export declare function registerTmuxBindings(): Promise<void>;
  * or user scripts may still read them). When all consumers have been
  * migrated and a major version ships, remove the `COMPOUND_*` lines.
  */
-export declare function buildEnv(cwd: string, v1SessionId?: string): Record<string, string>;
+export declare function buildEnv(cwd: string, v1SessionId?: string, runtime?: RuntimeHost): Record<string, string>;
 /**
  * Test-only exports for the C5 rendering pipeline. The ergonomic choice
  * over `export function normalizeDescription` is intentional: anything

package/dist/core/config-injector.js

@@ -428,12 +428,13 @@ export async function registerTmuxBindings() {
  * or user scripts may still read them). When all consumers have been
  * migrated and a major version ships, remove the `COMPOUND_*` lines.
  */
-export function buildEnv(cwd, v1SessionId) {
+export function buildEnv(cwd, v1SessionId, runtime = 'claude') {
     const env = {
         // New canonical names
         FORGEN_HARNESS: '1',
         FORGEN_CWD: cwd,
         FORGEN_V1: '1',
+        FORGEN_RUNTIME: runtime,
         // Legacy compat (remove in next major)
         COMPOUND_HARNESS: '1',
         COMPOUND_CWD: cwd,

package/dist/core/dashboard.d.ts

@@ -0,0 +1,108 @@
+/**
+ * Forgen — Compound Dashboard
+ *
+ * Provides a rich terminal overview of the compound knowledge system:
+ * knowledge inventory, injection activity, lifecycle transitions,
+ * session history, and hook health.
+ *
+ * Data is collected from:
+ *   - ME_SOLUTIONS, ME_RULES, ME_BEHAVIOR (knowledge files)
+ *   - MATCH_EVAL_LOG_PATH (injection/matching decisions)
+ *   - STATE_DIR (hook-errors.json, last-extraction.json)
+ *   - Solution frontmatter (lifecycle evidence fields)
+ */
+import type { SolutionStatus } from '../engine/solution-format.js';
+export interface KnowledgeOverview {
+    solutions: {
+        total: number;
+        byStatus: Record<SolutionStatus, number>;
+    };
+    rules: {
+        total: number;
+        categories: Record<string, number>;
+    };
+    behavior: {
+        total: number;
+    };
+    dateRange: {
+        oldest: string | null;
+        newest: string | null;
+    };
+}
+export interface InjectionActivity {
+    totalRecords: number;
+    recentInjections: Array<{
+        name: string;
+        ts: string;
+        source: string;
+    }>;
+    topSolutions: Array<{
+        name: string;
+        count: number;
+    }>;
+    hookCount: number;
+    mcpCount: number;
+}
+export interface ReflectionData {
+    totalSolutions: number;
+    reflectedCount: number;
+    unreflectedCount: number;
+    reflectionRate: number;
+}
+export interface LifecycleActivity {
+    recentPromotionCandidates: Array<{
+        name: string;
+        status: SolutionStatus;
+        evidence: {
+            reflected: number;
+            sessions: number;
+            negative: number;
+        };
+    }>;
+    statusDistribution: Record<SolutionStatus, number>;
+}
+export interface SessionHistory {
+    lastExtraction: {
+        date: string;
+        extractionsToday: number;
+    } | null;
+}
+export interface HookHealth {
+    errors: Array<{
+        hookName: string;
+        count: number;
+        lastAt: string;
+    }>;
+}
+/** Collect knowledge overview data. */
+export declare function collectKnowledgeOverview(): KnowledgeOverview;
+/** Collect injection activity from match-eval-log. */
+export declare function collectInjectionActivity(): InjectionActivity;
+/** Collect code reflection data from solution evidence. */
+export declare function collectReflectionData(): ReflectionData;
+/** Collect lifecycle activity data. */
+export declare function collectLifecycleActivity(): LifecycleActivity;
+/** Collect session extraction history. */
+export declare function collectSessionHistory(): SessionHistory;
+/** Collect hook error data. */
+export declare function collectHookHealth(): HookHealth;
+export interface LearningCurve {
+    correctionsLast7d: number;
+    correctionsPrev7d: number;
+    correctionTrend: 'improving' | 'stable' | 'worsening';
+    evidenceTotalDays: number;
+    sessionsAnalyzed: number;
+    estimatedMinutesSaved: number;
+    topCorrectionAxes: Array<{
+        axis: string;
+        count: number;
+    }>;
+}
+/**
+ * Learning Curve 수집.
+ * evidence 파일(교정 기록)과 compound 활용률을 교차 분석하여 "쓸수록 나아진다"를 정량화.
+ */
+export declare function collectLearningCurve(): LearningCurve;
+export declare function renderDashboard(): string;
+/** CLI handler: forgen dashboard */
+export declare function handleDashboard(): Promise<void>;