oh-my-knowledge 1.0.0-beta.6 → 1.0.0-beta.7

package/dist/assets/agent-skills/omk/references/commands.md

@@ -73,7 +73,7 @@ omk eval [flags]
 
 **Flags:**
 
-- `--batch` `boolean`:batch 模式：baseline 作为对照组，逐个 skill 作为实验组
+- `--batch` `boolean`:批量评测：baseline 作为对照组，逐个 skill 作为实验组；不支持 repeat 大于 1
 - `--bootstrap` `boolean`:加 bootstrap CI
 - `--bootstrap-samples` `option`:bootstrap 重采样次数，默认 1000
 - `--budget-per-sample-ms` `option`:单用例时长上限 ms（必须 > 0，不传则无上限）
@@ -95,7 +95,6 @@ omk eval [flags]
 - `--layered-stats` `boolean`:输出分层统计
 - `--mcp-config` `option`:MCP 配置文件路径
 - `--model` `option`:被测模型
-- `--no-cache` `boolean`:跳过 executor cache
 - `--no-debias-length` `boolean`:关 length-debias（默认开）
 - `--no-diagnostic` `boolean`:关闭基于 Core 失败、缺失、排除与稳定 reason code 的诊断投影。
 - `--no-evidence` `boolean`:不把本次评测写成证据追加进受管记录(默认会为已 install 的 skill 自动写)。

package/dist/cli/commands/eval/gold/index.js

@@ -1,11 +1,5 @@
-// oclif 版 eval gold topic shim — 裸 `omk eval gold` 走 oclif Help class 打 topic
-// help,然后 exit 1(跟 legacy CliExit(1) 行为对齐:CI 脚本靠 exit code 区分「漏写
-// sub-sub」)。
-//
-// 不再依赖 legacy hand-written usage() 字符串(已删):oclif 默认没内建 help command,
-// 但 helpClass 是可实例化的 Help 子类。直接 new HelpClass(config).showCommandHelp(
-// config.findCommand('eval:gold')) 让 LangAwareHelp 按 --lang 渲染当前 topic 的
-// description + sub-sub 列表(init / validate / compare)。
+// 裸 `omk eval gold` 显示帮助并退出 1，让自动化能识别缺少子命令。
+// 复用当前 HelpClass，保持语言选择和子命令列表与 CLI 一致。
 import { BaseCommand } from '../../../oclif/base-command.js';
 import { LANG_FLAG, bilingual } from '../../../oclif/i18n.js';
 export default class EvalGold extends BaseCommand {

package/dist/cli/commands/eval/index.d.ts

@@ -1,7 +1,15 @@
 import { BaseCommand } from '../../oclif/base-command.js';
 import { type CliLang } from '../../lib/i18n.js';
-import { type RunConfig } from '../../lib/parse-run-config.js';
-export declare function formatConnectivityFailureHint(message: string, config: Pick<RunConfig, 'executorName' | 'model' | 'judgeModels' | 'noJudge'>, lang: CliLang, env?: NodeJS.ProcessEnv): string;
+interface ConnectivityFailureContext {
+    executorName: string;
+    model: string;
+    judgeModels: readonly {
+        executor: string;
+        model: string;
+    }[];
+    noJudge: boolean;
+}
+export declare function formatConnectivityFailureHint(message: string, config: Pick<ConnectivityFailureContext, 'executorName' | 'model' | 'judgeModels' | 'noJudge'>, lang: CliLang, env?: NodeJS.ProcessEnv): string;
 export default class Eval extends BaseCommand {
     static description: string;
     static examples: {
@@ -23,7 +31,6 @@ export default class Eval extends BaseCommand {
         'output-dir': import("@oclif/core/interfaces").OptionFlag<string | undefined, import("@oclif/core/interfaces").CustomOptions>;
         global: import("@oclif/core/interfaces").BooleanFlag<boolean>;
         'no-judge': import("@oclif/core/interfaces").BooleanFlag<boolean>;
-        'no-cache': import("@oclif/core/interfaces").BooleanFlag<boolean>;
         'dry-run': import("@oclif/core/interfaces").BooleanFlag<boolean>;
         concurrency: import("@oclif/core/interfaces").OptionFlag<string | undefined, import("@oclif/core/interfaces").CustomOptions>;
         timeout: import("@oclif/core/interfaces").OptionFlag<string | undefined, import("@oclif/core/interfaces").CustomOptions>;
@@ -58,3 +65,4 @@ export default class Eval extends BaseCommand {
     };
     run(): Promise<void>;
 }
+export {};

package/dist/cli/commands/eval/index.js

@@ -1,12 +1,13 @@
 import { existsSync } from 'node:fs';
-import { join, relative, sep } from 'node:path';
+import { join, relative, resolve, sep } from 'node:path';
 import { Flags } from '@oclif/core';
 import { LANG_FLAG, bilingual } from '../../oclif/i18n.js';
 import { BaseCommand } from '../../oclif/base-command.js';
 import { enumStringParser, integerStringParser, numberStringParser } from '../../oclif/parsers.js';
 import { CliExit } from '../../lib/cli-exit.js';
 import { tCli } from '../../lib/i18n.js';
-import { parseRunConfig } from '../../lib/parse-run-config.js';
+import { prepareCliEvaluation } from '../../lib/prepare-evaluation.js';
+import { CliEvaluationInputError } from '../../../eval-workflows/hosts/application.js';
 import { codexModelFlagValue, codexModelHint } from '../../lib/codex-model-hint.js';
 import { looksLikeModelUnavailableFailure } from '../../lib/llm-failure-classifier.js';
 import { DEFAULT_EVALUATION_GATE_THRESHOLD as DEFAULT_GATE_THRESHOLD } from '../../../eval-workflows/evaluation-defaults.js';
@@ -15,7 +16,7 @@ import { shellQuoteArg } from '../../../shared/shell-quote.js';
 import { executorNamesForFamily } from '../../../executors/core/registry.js';
 import { DEFAULT_EVALUATION_TIMEOUT_MS } from '../../../eval-workflows/evaluation-defaults.js';
 // oclif 版 eval(默认 = run 模式) — 单次 typed parse 之后业务 inline。flag schema
-// 镜像 RUN_OPTIONS + eval-runner extra = 41 flag。具体语义跟约束在 parseRunConfig 里。
+// 参数语义与约束由 Workflow 的 parseCliEvaluationRequest 维护。
 //
 // `omk eval gold ...` 由 src/cli/commands/eval/gold/{init,validate,compare}.ts
 // 处理,oclif 文件目录路由自动接管,不进 eval.ts。
@@ -127,15 +128,29 @@ export function formatConnectivityFailureHint(message, config, lang, env = proce
     return '';
 }
 async function runEval(_args, flags, lang) {
-    const { values, config, evalConfig } = parseRunConfig({ ...flags }, { lang });
-    if (!values.batch && !hasUsableSamplesPath(config.samplesPath)) {
+    let prepared;
+    try {
+        prepared = prepareCliEvaluation({ ...flags }, { lang });
+    }
+    catch (error) {
+        if (!(error instanceof CliEvaluationInputError))
+            throw error;
+        console.error(`error: ${error.message}`);
+        throw new CliExit(2);
+    }
+    const { request } = prepared;
+    const values = flags;
+    const evalConfig = prepared.parseInput.evalConfig;
+    const samplesPath = resolve(prepared.projectRoot, request.values.locators.samples);
+    const skillDir = resolve(prepared.projectRoot, request.values.locators.skillDirectory);
+    if (!values.batch && !hasUsableSamplesPath(samplesPath)) {
         const treatmentRaw = typeof values.treatment === 'string' ? values.treatment : '';
         const treatments = treatmentRaw.split(',').map((v) => v.trim()).filter(Boolean);
         const sampleCommand = !values.samples && !evalConfig?.samples && treatments.length === 1
-            ? sampleCommandForSingleTreatment(treatments[0], config.skillDir)
+            ? sampleCommandForSingleTreatment(treatments[0], skillDir)
             : null;
         const missingSamplesMessage = [
-            tCli('cli.common.samples_not_found', lang, { path: config.samplesPath }),
+            tCli('cli.common.samples_not_found', lang, { path: samplesPath }),
             sampleCommand ? tCli('cli.common.samples_not_found_hint', lang, { command: sampleCommand }) : '',
         ].filter(Boolean).join('\n');
         console.error(tCli('cli.common.error_prefix', lang, {
@@ -145,12 +160,7 @@ async function runEval(_args, flags, lang) {
     }
     try {
         const { runCoreEvaluationCommand } = await import('../../lib/run-core-evaluation.js');
-        const result = await runCoreEvaluationCommand({
-            flags: { ...flags },
-            config,
-            evalConfig,
-            lang,
-        });
+        const result = await runCoreEvaluationCommand({ prepared });
         if (!process.stdout.isTTY || result.output && typeof result.output === 'object'
             && result.output.projectionKind === 'core-cli-dry-run') {
             console.log(JSON.stringify(result.output, null, 2));
@@ -171,7 +181,12 @@ async function runEval(_args, flags, lang) {
             throw err;
         const message = err instanceof Error ? err.message : String(err);
         console.error(tCli('cli.common.error_prefix', lang, {
-            message: `${message}${formatConnectivityFailureHint(message, config, lang)}`,
+            message: `${message}${formatConnectivityFailureHint(message, {
+                executorName: request.values.targetRuntime.executorId,
+                model: request.values.targetRuntime.model,
+                noJudge: !request.values.judges.enabled,
+                judgeModels: request.values.judges.members.map((member) => ({ executor: member.executorId, model: member.model })),
+            }, lang, prepared.environment.environment)}`,
         }));
         throw new CliExit(1);
     }
@@ -262,9 +277,6 @@ export default class Eval extends BaseCommand {
         'no-judge': Flags.boolean({
             description: bilingual({ zh: '跳过 LLM 评委', en: 'Skip LLM judge' }),
         }),
-        'no-cache': Flags.boolean({
-            description: bilingual({ zh: '跳过 executor cache', en: 'Skip executor cache' }),
-        }),
         'dry-run': Flags.boolean({
             description: bilingual({ zh: '只 plan 不实跑', en: 'Plan only, no real exec' }),
         }),
@@ -281,8 +293,8 @@ export default class Eval extends BaseCommand {
         }),
         batch: Flags.boolean({
             description: bilingual({
-                zh: 'batch 模式：baseline 作为对照组，逐个 skill 作为实验组',
-                en: 'Batch mode: baseline vs each skill',
+                zh: '批量评测：baseline 作为对照组，逐个 skill 作为实验组；不支持 repeat 大于 1',
+                en: 'Batch mode: baseline vs each skill; repeat must be 1',
             }),
         }),
         'skip-connectivity': Flags.boolean({

package/dist/cli/commands/evolve.js

@@ -92,7 +92,7 @@ export async function runEvolve(args, flags, lang) {
         samplesFile = resolvedInput.samplesPath;
     }
     // 参数校验必须早于任何昂贵副作用(自动生成用例 / LLM 调用)。
-    const { parseJudgeModelsArgOrExit } = await import('../lib/parse-run-config.js');
+    const { parseJudgeModelsArgOrExit } = await import('../lib/parse-run-config/judge-models.js');
     const evolveJudges = parseJudgeModelsArgOrExit(flags['judge-models']);
     if (evolveJudges.length > 1) {
         console.error(tCli('cli.common.judge_models_single_only', lang, { cmd: 'omk evolve' }));
@@ -168,13 +168,15 @@ export async function runEvolve(args, flags, lang) {
     }
     const { evolveSkillCore } = await import('../../knowledge-artifacts/authoring/core-evolver.js');
     const { runCoreEvaluationCommand } = await import('../lib/run-core-evaluation.js');
+    const { prepareCliEvaluation } = await import('../lib/prepare-evaluation.js');
     const evolveEffort = flags.effort ? validateEvolveEffort(flags.effort, lang) : undefined;
     const evaluatePair = async (control, treatment) => {
         const evaluation = await runCoreEvaluationCommand({
-            flags: {
+            prepared: prepareCliEvaluation({
                 control,
                 treatment,
                 samples: resolve(samplesFile),
+                'skill-dir': dirname(resolve(skillPath)),
                 executor: flags.executor,
                 model: flags.model,
                 'judge-models': evolveJudges.map((judge) => `${judge.executor}:${judge.model}`).join(','),
@@ -185,17 +187,7 @@ export async function runEvolve(args, flags, lang) {
                 'no-evidence': true,
                 'no-serve': true,
                 'report-only': true,
-            },
-            config: {
-                samplesPath: resolve(samplesFile),
-                skillDir: dirname(resolve(skillPath)),
-                executorName: flags.executor,
-                model: flags.model,
-                effort: evolveEffort,
-                judgeModels: evolveJudges,
-            },
-            evalConfig: null,
-            lang: 'zh',
+            }, { lang: 'zh' }),
         });
         if (evaluation.stored === undefined) {
             throw new Error('Core evolve evaluation 未持久化 artifact chain。');

package/dist/cli/commands/init.js

@@ -91,8 +91,7 @@ export default class Init extends BaseCommand {
             }),
             required: false,
             parse: async (input) => {
-                // 拒绝 `omk init -- --weird` 这种把 flag 当 positional 的写法 — legacy 会
-                // 创建名为 `--weird` 的目录,新人一头雾水。在 oclif Args 这层拦住更友好。
+                // 在参数解析阶段拒绝疑似遗漏的 flag，避免误建名为 `--weird` 的目录。
                 if (input.startsWith('--')) {
                     const lang = resolveLang();
                     const msg = lang === 'zh'

package/dist/cli/commands/observe/inbox.js

@@ -114,7 +114,9 @@ export async function runObserveInbox(_args, flags, lang) {
         items = items.slice(0, limit);
     }
     if (flags.json) {
-        console.log(JSON.stringify({ kind: 'observe-inbox-query', items }, null, 2));
+        const { buildObservationInboxViewModel } = await import('../../../observability/inbox/view-model.js');
+        const { effectiveExperienceReports, resolvedReviewSessions, unappliedMetricAnnotations } = buildObservationInboxViewModel(dir, { skill: flags.skill });
+        console.log(JSON.stringify({ kind: 'observe-inbox-query', items, effectiveExperienceReports, resolvedReviewSessions, unappliedMetricAnnotations }, null, 2));
         return;
     }
     if (items.length === 0) {

package/dist/cli/commands/promote.js

@@ -1,4 +1,4 @@
-import { execFileSync } from 'node:child_process';
+import { resolveActor } from '../lib/actor.js';
 import { Args, Flags } from '@oclif/core';
 import { LANG_FLAG, bilingual } from '../oclif/i18n.js';
 import { BaseCommand } from '../oclif/base-command.js';
@@ -7,18 +7,6 @@ import { CliExit } from '../lib/cli-exit.js';
 import { sanitizeCell } from '../lib/cell-format.js';
 import { appendManagedDecision, evaluatePromoteGate, globalManagedDir, isCurrentlyPromoted, loadManagedRecord, probeSourceState, managedDir, managedRecordId, resolveManagedDir, } from '../../knowledge-artifacts/governance/index.js';
 const SUPPORTED_KINDS = ['skill'];
-/** 决定的 actor:--actor > git config user.name > $USER / $LOGNAME > unknown。git 参数无用户输入,安全。 */
-function resolveActor(flagActor) {
-    if (flagActor && flagActor.trim())
-        return flagActor.trim();
-    try {
-        const name = execFileSync('git', ['config', 'user.name'], { encoding: 'utf-8', stdio: ['ignore', 'pipe', 'ignore'] }).trim();
-        if (name)
-            return name;
-    }
-    catch { /* git 缺失 / 无配置 → 回退环境 */ }
-    return process.env.USER || process.env.LOGNAME || 'unknown';
-}
 /** 洗 detail 里来自受管 JSON 的不可信值(verdict / judgePromptHash),文本输出防 ANSI/OSC 注入。 */
 function safeDetail(detail) {
     if (!detail)

package/dist/cli/commands/rollback.js

@@ -1,4 +1,4 @@
-import { execFileSync } from 'node:child_process';
+import { resolveActor } from '../lib/actor.js';
 import { Args, Flags } from '@oclif/core';
 import { LANG_FLAG, bilingual } from '../oclif/i18n.js';
 import { BaseCommand } from '../oclif/base-command.js';
@@ -7,18 +7,6 @@ import { CliExit } from '../lib/cli-exit.js';
 import { sanitizeCell } from '../lib/cell-format.js';
 import { appendManagedDecision, globalManagedDir, isCurrentlyPromoted, loadManagedRecord, managedDir, managedRecordId, resolveManagedDir, } from '../../knowledge-artifacts/governance/index.js';
 const SUPPORTED_KINDS = ['skill'];
-/** 决定的 actor:--actor > git config user.name > $USER / $LOGNAME > unknown。git 参数无用户输入,安全。 */
-function resolveActor(flagActor) {
-    if (flagActor && flagActor.trim())
-        return flagActor.trim();
-    try {
-        const name = execFileSync('git', ['config', 'user.name'], { encoding: 'utf-8', stdio: ['ignore', 'pipe', 'ignore'] }).trim();
-        if (name)
-            return name;
-    }
-    catch { /* git 缺失 / 无配置 → 回退环境 */ }
-    return process.env.USER || process.env.LOGNAME || 'unknown';
-}
 export default class Rollback extends BaseCommand {
     static description = bilingual({
         zh: '回退受管 skill 当前版本的 promoted 接受：撤销最近一次 promote，在记录里追加一条 rollback 决定（源未漂移则状态回到 measurable，源已漂移则仍 stale）。',

package/dist/cli/lib/actor.d.ts

@@ -0,0 +1,2 @@
+/** Resolve a decision actor from the flag, Git config, environment, then fallback. */
+export declare function resolveActor(flagActor: string | undefined): string;

package/dist/cli/lib/actor.js

@@ -0,0 +1,13 @@
+import { execFileSync } from 'node:child_process';
+/** Resolve a decision actor from the flag, Git config, environment, then fallback. */
+export function resolveActor(flagActor) {
+    if (flagActor && flagActor.trim())
+        return flagActor.trim();
+    try {
+        const name = execFileSync('git', ['config', 'user.name'], { encoding: 'utf-8', stdio: ['ignore', 'pipe', 'ignore'] }).trim();
+        if (name)
+            return name;
+    }
+    catch { /* git 缺失 / 无配置 → 回退环境 */ }
+    return process.env.USER || process.env.LOGNAME || 'unknown';
+}

package/dist/cli/lib/cmd-flags.d.ts

@@ -36,7 +36,7 @@
  *   优先级 `--lang CLI flag > OMK_LANG env > zh`。
  *
  * 三态 boolean 注意:oclif 对 `Flags.boolean({})`(没 `default`)运行时不会塞 `false`,
- * 而是字段缺失,字段读出来是 `undefined`。这点对 `eval` 跟 `parseRunConfig()` fallback
+ * 而是字段缺失,字段读出来是 `undefined`。这点对 `eval` 跟 `parseCliEvaluationRequest()` fallback
  * 关键 — `undefined`(CLI 没传)走 `eval.yaml` config / 内置 default;`false`(显式
  * `--no-xxx`)真的关。如果在 interface 里把它标成 `boolean`,后续维护者按类型读
  * `flags.bootstrap` 会静默把「没传」当成「关」,绕开 config fallback。所以未设
@@ -80,7 +80,6 @@ export interface EvalFlags {
     /** 报告写全局而非项目 .omk/eval(写入侧 escape;读取侧 studio / 复用走 overlay 兜底)。 */
     global?: boolean;
     'no-judge'?: boolean;
-    'no-cache'?: boolean;
     'dry-run'?: boolean;
     concurrency?: string;
     timeout?: string;

package/dist/cli/lib/cmd-flags.js

@@ -36,7 +36,7 @@
  *   优先级 `--lang CLI flag > OMK_LANG env > zh`。
  *
  * 三态 boolean 注意:oclif 对 `Flags.boolean({})`(没 `default`)运行时不会塞 `false`,
- * 而是字段缺失,字段读出来是 `undefined`。这点对 `eval` 跟 `parseRunConfig()` fallback
+ * 而是字段缺失,字段读出来是 `undefined`。这点对 `eval` 跟 `parseCliEvaluationRequest()` fallback
  * 关键 — `undefined`(CLI 没传)走 `eval.yaml` config / 内置 default;`false`(显式
  * `--no-xxx`)真的关。如果在 interface 里把它标成 `boolean`,后续维护者按类型读
  * `flags.bootstrap` 会静默把「没传」当成「关」,绕开 config fallback。所以未设

package/dist/cli/lib/i18n-dict/common.d.ts

@@ -1,3 +1,3 @@
 import type { CliMessage } from './types.js';
-export type CommonMessageKey = 'cli.common.unknown_domain' | 'cli.common.error_prefix' | 'cli.common.skill_dir_not_found' | 'cli.common.skill_file_not_found' | 'cli.common.skill_dir_no_skill_md' | 'cli.common.report_not_found' | 'cli.common.no_judge_model' | 'cli.common.judge_models_single_only' | 'cli.common.warn_load_samples_failed' | 'cli.common.ambiguous_sample_files' | 'cli.common.samples_not_found' | 'cli.common.samples_not_found_hint' | 'cli.update.new_version_available' | 'cli.update.box_title' | 'cli.update.box_version_line' | 'cli.update.box_upgrade_line' | 'cli.update.box_silence_line' | 'cli.observe.view_hint' | 'cli.observe.observation_recorded' | 'cli.observe.production_gap' | 'cli.studio.started' | 'cli.studio.stop_hint' | 'cli.studio.open_failed' | 'cli.doctor.no_skill_found' | 'cli.doctor.progress_skill_start' | 'cli.doctor.progress_skill_done';
+export type CommonMessageKey = 'cli.common.unknown_domain' | 'cli.common.error_prefix' | 'cli.common.skill_dir_not_found' | 'cli.common.skill_file_not_found' | 'cli.common.skill_dir_no_skill_md' | 'cli.common.judge_models_single_only' | 'cli.common.ambiguous_sample_files' | 'cli.common.samples_not_found' | 'cli.common.samples_not_found_hint' | 'cli.update.new_version_available' | 'cli.update.box_title' | 'cli.update.box_version_line' | 'cli.update.box_upgrade_line' | 'cli.update.box_silence_line' | 'cli.observe.view_hint' | 'cli.observe.observation_recorded' | 'cli.observe.production_gap' | 'cli.studio.started' | 'cli.studio.stop_hint' | 'cli.studio.open_failed' | 'cli.doctor.no_skill_found' | 'cli.doctor.progress_skill_start' | 'cli.doctor.progress_skill_done';
 export declare const commonDict: Record<CommonMessageKey, CliMessage>;

package/dist/cli/lib/i18n-dict/common.js

@@ -19,22 +19,10 @@ export const commonDict = {
         zh: '目录下未找到 SKILL.md: {path}',
         en: 'SKILL.md not found in directory: {path}',
     },
-    'cli.common.report_not_found': {
-        zh: '未找到 report: {id}',
-        en: 'Report not found: {id}',
-    },
-    'cli.common.no_judge_model': {
-        zh: '未指定评委。请加 --judge-models <executor:model>, 或确保 report.meta.judgeModels 已写。',
-        en: 'No judge configured. Pass --judge-models <executor:model> or ensure the report has meta.judgeModels.',
-    },
     'cli.common.judge_models_single_only': {
         zh: '{cmd} 仅支持单评委。--judge-models 只能传一个 executor:model entry。',
         en: '{cmd} only supports a single judge. --judge-models accepts exactly one executor:model entry.',
     },
-    'cli.common.warn_load_samples_failed': {
-        zh: '⚠ 加载 samples 文件失败 ({path}): {message}\n',
-        en: '⚠ Failed to load samples file ({path}): {message}\n',
-    },
     'cli.common.ambiguous_sample_files': {
         zh: '评测用例文件存在歧义：{paths}。同一作用域只能保留 eval-samples.json 或 eval-samples.yaml 其中一个。',
         en: 'Ambiguous eval sample files: {paths}. Keep only one of eval-samples.json or eval-samples.yaml in the same scope.',

package/dist/cli/lib/i18n-dict/help.d.ts

@@ -1,3 +1,3 @@
 import type { CliMessage } from './types.js';
-export type HelpMessageKey = 'cli.help.observe' | 'cli.help.observe_ingest' | 'cli.help.observe_inbox' | 'cli.help.observe_show' | 'cli.help.evolve' | 'cli.help.sample' | 'cli.help.studio';
+export type HelpMessageKey = 'cli.help.observe';
 export declare const helpDict: Record<HelpMessageKey, CliMessage>;

package/dist/cli/lib/i18n-dict/help.js

@@ -33,232 +33,6 @@ Options:
   --global                            Write to global ~/.oh-my-knowledge/observe/health instead of project .omk/observe/health
 
 The observe inbox is a separate line; see omk observe inbox --help.
-`,
-    },
-    'cli.help.observe_ingest': {
-        zh: `
-omk observe ingest — 读取真实 session trace，写入 observe inbox 数据
-
-用法：
-  omk observe ingest <sessions-dir-or-file> [options]
-
-选项：
-  --output-dir <path>                输出目录（默认：.omk/observe/inbox；读取时兜底到 ~/.oh-my-knowledge/observe/inbox）
-
-支持：
-  Claude Code JSONL
-  Markdown 对话日志（.log）
-`,
-        en: `
-omk observe ingest — read real session traces and write observe inbox data
-
-Usage:
-  omk observe ingest <sessions-dir-or-file> [options]
-
-Options:
-  --output-dir <path>                Output directory (default: .omk/observe/inbox; read fallback: ~/.oh-my-knowledge/observe/inbox)
-
-Supported:
-  Claude Code JSONL
-  Markdown conversation logs (.log)
-`,
-    },
-    'cli.help.observe_inbox': {
-        zh: `
-omk observe inbox — 查看已写入的 observe inbox 问题列表
-
-用法：
-  omk observe inbox [options]
-
-选项：
-  --input-dir <path>                 读取目录（默认：.omk/observe/inbox；兜底到 ~/.oh-my-knowledge/observe/inbox）
-  --limit <n>                        展示 top N（默认：20）
-  --skill <name>                     只看指定 skill
-  --explore <n>                      从最近 50 条 medium/low 问题里抽样查看长尾
-  --include-noise                    --explore 时显式包含 noise 桶
-  --by-skill                         按 skill 输出资产看板
-  --llm-enhanced-review              显式调用模型进行链路增强复盘
-  --model <name>                     LLM 增强复盘使用的模型（默认：sonnet）
-  --executor <name>                  LLM 增强复盘使用的执行器
-  --refresh                          强制重新生成 LLM 增强复盘
-  --json                             输出 JSON
-`,
-        en: `
-omk observe inbox — inspect previously ingested observe inbox items
-
-Usage:
-  omk observe inbox [options]
-
-Options:
-  --input-dir <path>                 Input directory (default: .omk/observe/inbox; fallback: ~/.oh-my-knowledge/observe/inbox)
-  --limit <n>                        Show top N (default: 20)
-  --skill <name>                     Only show one skill
-  --explore <n>                      Sample long-tail issues from the latest 50 medium/low items
-  --include-noise                    Explicitly include the noise bucket with --explore
-  --by-skill                         Print the skill-level asset board
-  --llm-enhanced-review              Explicitly run model-based enhanced chain review
-  --model <name>                     Model for LLM enhanced review (default: sonnet)
-  --executor <name>                  Executor for LLM enhanced review
-  --refresh                          Force LLM enhanced review refresh
-  --json                             Print JSON
-`,
-    },
-    'cli.help.observe_show': {
-        zh: `
-omk observe show — 查看单条 observation 的原始上下文
-
-用法：
-  omk observe show <inbox_id> [options]
-
-选项：
-  --input-dir <path>                 读取目录（默认：.omk/observe/inbox；兜底到 ~/.oh-my-knowledge/observe/inbox）
-`,
-        en: `
-omk observe show — inspect the raw context around one observation
-
-Usage:
-  omk observe show <inbox_id> [options]
-
-Options:
-  --input-dir <path>                 Input directory (default: .omk/observe/inbox; fallback: ~/.oh-my-knowledge/observe/inbox)
-`,
-    },
-    'cli.help.evolve': {
-        zh: `
-omk evolve——多轮自动迭代改进 skill
-
-用法：
-  omk evolve <skill-path> [options]
-
-选项：
-  --rounds <n>                        迭代轮数（默认：5）
-  --target <score>                    目标分数
-  --model <name>                      任务执行模型，默认跟随 runtime；Codex 读取本机配置
-  --improve-model <name>              skill 改写模型，默认沿用任务执行模型
-  --judge-models <executor:model>     单评委配置（默认跟随执行器；Codex 沿用被测模型）
-
-示例：
-  omk evolve skills/code-review/SKILL.md
-  omk evolve skills/code-review/SKILL.md --rounds 10 --target 4.5
-  omk evolve skills/code-review/SKILL.md --model sonnet --improve-model opus
-`,
-        en: `
-omk evolve — auto-iterate a skill through multi-round evaluation loops
-
-Usage:
-  omk evolve <skill-path> [options]
-
-Options:
-  --rounds <n>                        Iteration rounds (default: 5)
-  --target <score>                    Target score
-  --model <name>                      Task executor model; follows runtime (Codex reads local config)
-  --improve-model <name>              Skill rewriter model; defaults to the task model
-  --judge-models <executor:model>     Single judge config (follows executor; Codex reuses task model)
-
-Examples:
-  omk evolve skills/code-review/SKILL.md
-  omk evolve skills/code-review/SKILL.md --rounds 10 --target 4.5
-  omk evolve skills/code-review/SKILL.md --model sonnet --improve-model opus
-`,
-    },
-    'cli.help.sample': {
-        zh: `
-omk sample——生成或补齐 eval-samples 评测用例
-
-用法：
-  omk sample <skill-path> [options]
-  omk sample --batch [--skill-dir <dir>] [options]
-
-输出位置（默认）：
-  目录 skill（<skill>/SKILL.md） → <skill>/.omk/eval-samples.json（也支持 .yaml）
-  扁平 .md（单次）               → 当前目录的 eval-samples.json（项目级兜底）
-  扁平 .md（--batch）            → 跳过；私有用例请迁移为目录 skill
-
-选项：
-  --count <n>                         强制生成 N 条（不指定时由 LLM 按 skill 类型自动判断：
-                                       工作流型 6-8 条 / 原子型 4-6 条 / 混合型 5-7 条）
-  --model <name>                      生成模型（默认：opus；lean+effort-low 已自动开,想省钱可改 sonnet/haiku）
-  --focus <text>                      自然语言指定希望覆盖的场景（追加到 prompt，优先级高于自由发挥）
-  --batch                             为 skill 目录下缺少 samples 的 skill 批量生成
-  --skill-dir <path>                  skill 目录（batch 使用，默认：skills）
-
-示例：
-  omk sample skills/req-tool.md
-  omk sample skills/req-tool.md --count 8 \\
-    --focus "重点覆盖 tag 查询走 PROJECT 空 → WORKSPACE 兜底的多步流程，以及 search 失败的错误路径"
-`,
-        en: `
-omk sample — generate or fill eval-samples test cases
-
-Usage:
-  omk sample <skill-path> [options]
-  omk sample --batch [--skill-dir <dir>] [options]
-
-Output path (default):
-  directory skill (<skill>/SKILL.md) → <skill>/.omk/eval-samples.json (.yaml also supported)
-  flat .md (single)                  → ./eval-samples.json in current directory (project fallback)
-  flat .md (--batch)                 → skipped; migrate to a directory skill for private samples
-
-Options:
-  --count <n>                         Force N samples (omit to let LLM auto-decide by skill type:
-                                       workflow 6-8 / atomic 4-6 / mixed 5-7)
-  --model <name>                      Generation model (default: opus; lean+effort-low applied; pass --model sonnet/haiku to save cost)
-  --focus <text>                      Natural-language scenario hints appended to the prompt (overrides freeform diversity)
-  --batch                             Generate for skills that are missing samples
-  --skill-dir <path>                  Skill directory for batch mode (default: skills)
-
-Examples:
-  omk sample skills/req-tool.md
-  omk sample skills/req-tool.md --count 8 \\
-    --focus "Cover PROJECT-empty → WORKSPACE-fallback multi-step tag lookup and the search-failure error path"
-`,
-    },
-    'cli.help.studio': {
-        zh: `
-omk studio——打开本地知识工作台
-
-用法：
-  omk studio [options]
-
-选项：
-  --port <n>                          本地服务端口（默认：7799）
-  --host <host>                       监听地址（默认：127.0.0.1；局域网访问可用 0.0.0.0）
-  --reports-dir <path>                只看指定 Core 报告目录（默认聚合当前项目 + 全局）
-  --analyses-dir <path>               观测健康报告目录（默认：项目级 .omk/observe/health，空则全局兜底）
-  --doctors-dir <path>                体检报告目录（默认：项目级 .omk/doctor，空则全局兜底）
-  --observations-dir <path>           observe inbox 数据目录（默认：.omk/observe/inbox）
-  --global                            只看全局 eval / observe/health / doctor / observe/inbox 目录（~/.oh-my-knowledge/），而非机器级聚合 / 项目优先；governance/managed 不受影响
-  --no-open                           只启动服务，不自动打开浏览器
-  --dev                               开发模式：文件变化时自动重启
-
-示例：
-  omk studio
-  omk studio --port 7798
-  omk studio --host 0.0.0.0 --observations-dir .omk/observe/inbox
-  omk studio --no-open
-`,
-        en: `
-omk studio — open the local knowledge workbench
-
-Usage:
-  omk studio [options]
-
-Options:
-  --port <n>                          Local server port (default: 7799)
-  --host <host>                       Listen address (default: 127.0.0.1; use 0.0.0.0 for LAN access)
-  --reports-dir <path>                View only this Core reports dir (default aggregates current project + global)
-  --analyses-dir <path>               Observe-health reports dir (default: project .omk/observe/health, falls back to global)
-  --doctors-dir <path>                Doctor reports dir (default: project .omk/doctor, falls back to global)
-  --observations-dir <path>           Observe inbox data directory (default: .omk/observe/inbox)
-  --global                            View only global eval, observe/health, doctor, and observe/inbox directories under ~/.oh-my-knowledge/; does not affect governance/managed
-  --no-open                           Start the server without opening a browser
-  --dev                               Dev mode: restart on file changes
-
-Examples:
-  omk studio
-  omk studio --port 7798
-  omk studio --host 0.0.0.0 --observations-dir .omk/observe/inbox
-  omk studio --no-open
 `,
     },
 };

package/dist/cli/lib/i18n-dict/run.d.ts

@@ -1,3 +1,3 @@
 import type { CliMessage } from './types.js';
-export type RunMessageKey = 'cli.progress.preflight_starting' | 'cli.progress.sample_retry' | 'cli.progress.sample_error' | 'cli.progress.sample_executing' | 'cli.progress.sample_exec_done' | 'cli.progress.output_preview' | 'cli.progress.judging' | 'cli.progress.judged' | 'cli.progress.skipped' | 'cli.progress.sample_done' | 'cli.progress.sample_failed_done' | 'cli.run.invalid_repeat' | 'cli.run.invalid_holdout_ratio' | 'cli.run.invalid_judge_repeat' | 'cli.run.no_debias_length_active' | 'cli.run.invalid_bootstrap_samples' | 'cli.run.bootstrap_samples_too_large' | 'cli.run.dry_run_no_scores' | 'cli.run.skill_section' | 'cli.run.run_section' | 'cli.run.batch_complete' | 'cli.run.batch_verdict_header' | 'cli.run.batch_verdict_next_step' | 'cli.run.batch_child_report_missing' | 'cli.run.eval_complete' | 'cli.run.tally' | 'cli.run.report_saved' | 'cli.run.evidence_recorded' | 'cli.run.evidence_recorded_promotable' | 'cli.run.evidence_recorded_unbound' | 'cli.run.report_only_gate_skipped' | 'cli.run.report_server_running' | 'cli.run.report_server_view' | 'cli.run.report_server_stop' | 'cli.run.no_serve_in_non_tty' | 'cli.run.no_serve_view_hint' | 'cli.run.gold_load_failed' | 'cli.run.gold_load_issue' | 'cli.run.contamination_warning' | 'cli.run.codex_fallback_hint' | 'cli.run.codex_auth_hint' | 'cli.run.codex_model_hint' | 'cli.run.openai_api_auth_hint' | 'cli.run.openai_api_model_hint' | 'cli.run.anthropic_api_auth_hint' | 'cli.run.anthropic_api_model_hint' | 'cli.run.skip_connectivity_warning';
+export type RunMessageKey = 'cli.run.batch_verdict_header' | 'cli.run.codex_fallback_hint' | 'cli.run.codex_auth_hint' | 'cli.run.codex_model_hint' | 'cli.run.openai_api_auth_hint' | 'cli.run.openai_api_model_hint' | 'cli.run.anthropic_api_auth_hint' | 'cli.run.anthropic_api_model_hint';
 export declare const runDict: Record<RunMessageKey, CliMessage>;