oh-my-knowledge 0.45.0 → 0.47.0

package/README.md

@@ -8,8 +8,8 @@
 
 **English** | [简体中文](./README.zh.md)
 
-**Did your prompt actually get better?**
-A/B test your prompts and skills with statistical rigor — bootstrap CI and length-debias on by default, Krippendorff α the moment you add a gold set.
+**Is this knowledge input (prompt / skill / RAG / agent) any good, and can you ship it with evidence?**
+`doctor` checks whether this knowledge input is coherent enough to measure; `eval` fixes the model and samples, changes only the knowledge input, and tells you whether the new version is genuinely better. Bootstrap CI and length-debias are on by default; Krippendorff α appears the moment you add a gold set.
 
 📖 **Full documentation: [oh-my-knowledge.pages.dev](https://oh-my-knowledge.pages.dev)** (searchable, English / 简体中文)
 
@@ -35,6 +35,20 @@ Walkthrough: [5-minute quickstart guide](docs/quickstart-skill-eval.md) (recomme
 
 Deeper: [who omk is for](docs/explanation/who-omk-is-for.md) · [CLI reference](docs/reference/cli.md) · [how it works](docs/explanation/architecture.md) · [eval sample format](docs/reference/eval-sample-format.md) · [executors](docs/reference/executors.md) · [artifact layout](docs/reference/artifact-layout.md)
 
+## The first workflow
+
+omk is primarily for authors and maintainers of LLM knowledge artifacts who need a release decision, not for passive end-users of a skill. The first workflow is deliberately small:
+
+```text
+change a skill / prompt / agent artifact
+→ run omk doctor to catch structure, dependency, and measurability problems
+→ run omk eval to compare against a baseline on the same samples
+→ read the report / Studio view for the next concrete fix
+→ decide ship / don't ship
+```
+
+`observe` is the later production-feedback loop: useful once real usage traces exist, but not required for omk's first value. The trunk is the pre-ship doctor → eval decision.
+
 ## Use inside AI Coding Agents
 
 Install the official omk Agent Skill to let your coding agent run omk workflows from natural language:
@@ -73,7 +87,7 @@ You can also describe the goal in natural language, such as "compare v1 vs v2" o
 
 ## Why this tool
 
-Teams doing knowledge engineering produce lots of knowledge artifacts (skills today, but also prompts, agents, workflows…). When someone asks "why is v2 better than v1", you need objective data instead of gut feeling. `oh-my-knowledge` solves this with controlled experiments: **same model, same test samples, only the knowledge artifact changes.**
+Teams doing knowledge engineering produce lots of knowledge artifacts (skills today, but also prompts, agents, workflows…). When someone asks "can we ship v2, and why?", you need objective data instead of gut feeling. `oh-my-knowledge` solves this with controlled experiments: **same model, same test samples, only the knowledge artifact changes.**
 
 ## Why omk over alternatives
 
@@ -102,7 +116,7 @@ RAG-specific evals: see RAGAS (separate niche, complementary to omk). Full compa
 | **30+ assertion types** | substring, regex, JSON Schema, ROUGE/BLEU/Levenshtein similarity, agent tool-call assertions, semantic similarity, custom JS |
 | **Statistical rigor** | Bootstrap CI / length-debias / saturation curve on by default; Krippendorff α auto-computed with a gold set. [Details →](docs/explanation/statistical-rigor.md) |
 | **RAG metrics** | `faithfulness` / `answer_relevancy` / `context_recall` — anti-hallucination + answer relevance + context coverage |
-| **LLM health audit** | `omk doctor` grades 7 builtin dimensions; `--static-only` runs offline without an LLM |
+| **LLM health audit** | `omk doctor` grades 7 builtin dimensions; repeats the audit (`--repeat`) and merges findings by k/n consensus |
 | **Production observability** | parse Claude Code session JSONL traces; measure per-skill failure rate / latency / cost / knowledge-gap signals |
 | **Knowledge-gap detection** | severity-weighted signals quantify risk exposure instead of claiming completeness |
 | **Construct-validity isolation** | `--strict-baseline` (default ON) cuts three contamination channels so baseline doesn't silently see the skill it's being compared against |

package/README.zh.md

@@ -8,8 +8,8 @@
 
 [English](./README.md) | **简体中文**
 
-**你改完 prompt，真的变好了吗？**
-用统计严谨性 A/B 测试你的 prompt 和 skill —— Bootstrap 置信区间、长度去偏默认开，配一份人工 gold 就自动算 Krippendorff α。
+**一段知识输入（prompt / skill / RAG / agent）好不好，能不能有证据地发布？**
+`doctor` 先检查这类知识输入是否清楚到值得测；`eval` 固定模型和用例，只改变知识输入，判断新版本是否真的更好。Bootstrap 置信区间、长度去偏默认开，配一份人工 gold 就自动算 Krippendorff α。
 
 📖 **完整文档：[oh-my-knowledge.pages.dev/zh](https://oh-my-knowledge.pages.dev/zh/)**（可搜索，可切换英文）
 
@@ -35,6 +35,20 @@ omk eval --control code-review-v1 --treatment code-review-v2
 
 深入：[为谁、解决什么](docs/zh/explanation/who-omk-is-for.md) · [CLI 参考](docs/zh/reference/cli.md) · [工作原理](docs/zh/explanation/architecture.md) · [评测用例格式](docs/zh/reference/eval-sample-format.md) · [执行器](docs/zh/reference/executors.md) · [artifact 布局](docs/zh/reference/artifact-layout.md)
 
+## 第一条工作流
+
+omk 主要给 LLM 知识载体的作者 / 维护者用，帮他们做发布判断；它不是给被动安装 skill 的普通使用者用的。第一条工作流应该很短：
+
+```text
+改了一个 skill / prompt / agent artifact
+→ 跑 omk doctor，先抓结构、依赖和可测性问题
+→ 跑 omk eval，在同一批用例上和 baseline 对比
+→ 看报告 / Studio，知道下一步具体该改哪里
+→ 决定发布 / 不发布
+```
+
+`observe` 是后续的生产反馈闭环：等真实使用 trace 存在后很有价值，但不是 omk 第一价值的前提。主干先是发布前的 doctor → eval 判断。
+
 ## 在 AI Coding Agent 中使用
 
 安装 omk 官方 Agent Skill 后，可以直接用自然语言让 coding agent 跑 omk 工作流：
@@ -73,7 +87,7 @@ omk sample skills/my-skill.md
 
 ## 为什么需要这个工具
 
-做知识工程的团队会产出大量知识载体（当前常见是 skill，也包括 prompt、agent、workflow 等）。当被问到「v2 比 v1 好在哪」时，需要客观数据而非主观判断。`oh-my-knowledge` 通过控制变量实验解决这个问题：**相同模型、相同评测用例，只改变知识载体。**
+做知识工程的团队会产出大量知识载体（当前常见是 skill，也包括 prompt、agent、workflow 等）。当被问到「v2 能不能发、为什么」时，需要客观数据而非主观判断。`oh-my-knowledge` 通过控制变量实验解决这个问题：**相同模型、相同评测用例，只改变知识载体。**
 
 ## 为什么选 omk
 
@@ -102,7 +116,7 @@ RAG 专项评测请看 RAGAS（独立 niche，跟 omk 互补）。完整对比
 | **30+ 种断言** | 包含子串、正则、JSON Schema、ROUGE/BLEU/Levenshtein 相似度、Agent 工具调用、语义相似度、自定义函数等 |
 | **统计严谨性** | Bootstrap CI / 长度去偏 / 饱和曲线默认开，Krippendorff α 提供 gold 集即自动计算。[详情 →](docs/zh/explanation/statistical-rigor.md) |
 | **RAG metrics** | `faithfulness` / `answer_relevancy` / `context_recall` 三 metric — 反幻觉 + 切题度 + context 覆盖 |
-| **LLM 健康度审计** | `omk doctor` 给 7 个内置维度独立打分；`--static-only` 可离线无 LLM 调用 |
+| **LLM 健康度审计** | `omk doctor` 给 7 个内置维度独立打分；重复采样（`--repeat`）+ k/n 共识归并 |
 | **线上 session 观测** | 解析 Claude Code session JSONL，测量各 skill 的失败率、耗时、token 成本、知识缺口信号 |
 | **知识缺口识别** | 严重度加权的信号量化风险敞口，不宣称完备性 |
 | **用例隔离 (construct validity)** | `--strict-baseline`（默认开）三堵 baseline 拿到被测 skill 的污染路径 |

package/dist/assets/agent-skills/omk/SKILL.md

@@ -121,11 +121,11 @@ omk sample --batch
 ### 体检 skill 写法
 
 ```bash
-# 全量体检（含 LLM-judge 维度）
+# 健康度审计（默认 --repeat 2 采样 + k/n 共识归并）
 omk doctor
 
-# 只跑静态检查不调 LLM
-omk doctor --static-only
+# 单次快检（不采样、不归并，最省）
+omk doctor --repeat 1
 
 # 针对单 skill
 omk doctor skills/my-skill.md

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

@@ -6,7 +6,7 @@
 
 ## omk doctor
 
-体检 omk 工作目录，检查 skill 配置 / 依赖 / executor 连通性。
+体检 omk 工作目录：先跑静态规则，再对 skill 做多维度 LLM 健康度审计（默认 --repeat 2 采样 + 共识归并）。
 
 **用法:**
 
@@ -20,6 +20,7 @@ omk doctor [target] [flags]
 
 **Flags:**
 
+- `--concurrency` `option`:多次采样的并发数。默认 = --repeat（全并行，各遍相互独立，压墙钟时间）。设 1 = 串行。成本不变，只抬高瞬时并发（rate-limit 敏感时调小）。
 - `--dimensions` `option`:自定义维度配置文件（YAML），追加到内置 7 维度之后。每条维度二选一：promptSection（走 LLM 体检）或 endpoint（POST skill 快照给接口判定）。注意：endpoint 会把 SKILL.md 全文 + 子文件发到该地址，仅对可信配置/可信地址启用。
 - `--effort` `option`:LLM 推理 effort：low / medium / high / xhigh / max。
 - `--executor` `option`:执行器名，默认 claude。指定为测试 fixture 路径可在测试里跑（同 omk doctor）。
@@ -30,25 +31,31 @@ omk doctor [target] [flags]
 - `--lang` `option` (默认 `zh`):输出语言 zh|en，优先级 CLI > OMK_LANG env > zh。
 - `--model` `option`:LLM model 名，默认 sonnet。
 - `--output-dir` `option`:报告输出目录，默认项目级 .omk/doctors（--global 写全局）。
-- `--samples` `option`:用例文件路径（.json/.yaml）。不传则按 target / cwd 顺序自动发现。
-- `--static-only` `boolean`:离线静态模式，只跑 4 条静态 rule(skill_readable / skill_metadata / dependencies_present / samples_contract_aligned），不调 LLM。
+- `--repeat` `option`:健康度体检重复采样次数（self-consistency）。默认 2：并行跑 2 遍、finding 取并集并用 LLM 聚类归并同根因、标注支持度 k/N，压低单次采样方差。设 1 = 单次快速体检（不采样、不归并，最省）。
+- `--static-only` `boolean`:只跑静态检测（不调 LLM、不读 samples.json）：skill 可读性 / frontmatter 合法性 / 正文引用的脚本·CLI·文件·env 是否存在。CI 无 LLM 凭证或断网时用。
 - `--timeout` `option`:单次 LLM 会话超时秒数，默认 600(10 分钟）。
 
 **示例:**
 
-> 默认模式跑 LLM 健康度审计(7 内置维度）。
+> 默认模式跑静态规则 + LLM 健康度审计（7 内置维度）。
 
 ```bash
 omk doctor
 ```
 
-> 离线静态模式，只跑 4 条静态 rule，不调 LLM,CI 无 LLM 凭证时用。
+> 单次快速体检（不采样、不归并，最省）。
+
+```bash
+omk doctor --repeat 1
+```
+
+> 只跑静态检测（不调 LLM、不读 samples）：结构 + 正文依赖检查。
 
 ```bash
 omk doctor --static-only
 ```
 
-> JSON 输出 + 写 HTML 报告，给 CI 抓 exit code 同时人看。
+> JSON 输出 + 静默 gate，给 CI 抓 exit code 同时人看。
 
 ```bash
 omk doctor --json --gate

package/dist/cli/commands/doctor.d.ts

@@ -14,14 +14,15 @@ export default class Doctor extends BaseCommand {
         gate: import("@oclif/core/interfaces").BooleanFlag<boolean>;
         executor: import("@oclif/core/interfaces").OptionFlag<string | undefined, import("@oclif/core/interfaces").CustomOptions>;
         model: import("@oclif/core/interfaces").OptionFlag<string | undefined, import("@oclif/core/interfaces").CustomOptions>;
-        samples: 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>;
         'output-dir': import("@oclif/core/interfaces").OptionFlag<string | undefined, import("@oclif/core/interfaces").CustomOptions>;
         global: import("@oclif/core/interfaces").BooleanFlag<boolean>;
         dimensions: import("@oclif/core/interfaces").OptionFlag<string | undefined, import("@oclif/core/interfaces").CustomOptions>;
-        'static-only': import("@oclif/core/interfaces").BooleanFlag<boolean>;
         fix: import("@oclif/core/interfaces").BooleanFlag<boolean>;
         effort: import("@oclif/core/interfaces").OptionFlag<string | undefined, import("@oclif/core/interfaces").CustomOptions>;
+        repeat: import("@oclif/core/interfaces").OptionFlag<string | undefined, import("@oclif/core/interfaces").CustomOptions>;
+        concurrency: import("@oclif/core/interfaces").OptionFlag<string | undefined, import("@oclif/core/interfaces").CustomOptions>;
+        'static-only': import("@oclif/core/interfaces").BooleanFlag<boolean>;
     };
     run(): Promise<void>;
 }

package/dist/cli/commands/doctor.js

@@ -3,7 +3,7 @@ import { join, resolve } from 'node:path';
 import { Args, Flags } from '@oclif/core';
 import { LANG_FLAG, bilingual } from '../oclif/i18n.js';
 import { BaseCommand } from '../oclif/base-command.js';
-import { numberStringParser } from '../oclif/parsers.js';
+import { enumStringParser, integerStringParser, numberStringParser } from '../oclif/parsers.js';
 import { CliExit } from '../lib/cli-exit.js';
 import { tCli } from '../lib/i18n.js';
 import { makeDoctorProgress } from '../lib/progress.js';
@@ -12,32 +12,38 @@ import { indexDoctorWrite, removeDoctorCard } from '../../eval-core/artifact-ind
 import { doctorReportFileStem, isReportFileName, reportFilePath } from '../../eval-core/artifact-file-names.js';
 import { migrateLegacyReportFiles } from '../../eval-core/report-file-migration.js';
 import { projectDoctorsDir, globalDoctorsDir } from '../../eval-core/measurement-dirs.js';
-import { findDoctorDeprecatedSamplesHint, findDoctorSamplesPath } from '../../inputs/sample-locator.js';
 import { persistDoctorGraphSidecars, removeDoctorGraphSidecars } from '../../artifact-graph/doctor.js';
 export default class Doctor extends BaseCommand {
     static description = bilingual({
-        zh: '体检 omk 工作目录，检查 skill 配置 / 依赖 / executor 连通性。',
-        en: 'Preflight health checks for omk workdir: skill config / deps / executor connectivity.',
+        zh: '体检 omk 工作目录：先跑静态规则，再对 skill 做多维度 LLM 健康度审计（默认 --repeat 2 采样 + 共识归并）。',
+        en: 'Preflight health checks for omk workdir: static rules plus multi-dimension LLM health audit of your skills (default --repeat 2 sampling + consensus merge).',
     });
     static examples = [
         {
             description: bilingual({
-                zh: '默认模式跑 LLM 健康度审计(7 内置维度）。',
-                en: 'Default mode runs LLM-driven health audit (7 built-in dimensions).',
+                zh: '默认模式跑静态规则 + LLM 健康度审计（7 内置维度）。',
+                en: 'Default mode runs static rules plus LLM-driven health audit (7 built-in dimensions).',
             }),
             command: '<%= config.bin %> doctor',
         },
         {
             description: bilingual({
-                zh: '离线静态模式，只跑 4 条静态 rule，不调 LLM,CI 无 LLM 凭证时用。',
-                en: 'Offline static mode, only 4 static rules, no LLM call. Use when CI lacks LLM credentials.',
+                zh: '单次快速体检（不采样、不归并，最省）。',
+                en: 'Single quick pass (no sampling/merge, cheapest).',
+            }),
+            command: '<%= config.bin %> doctor --repeat 1',
+        },
+        {
+            description: bilingual({
+                zh: '只跑静态检测（不调 LLM、不读 samples）：结构 + 正文依赖检查。',
+                en: 'Static checks only (no LLM, no samples): structural + body-dependency checks.',
             }),
             command: '<%= config.bin %> doctor --static-only',
         },
         {
             description: bilingual({
-                zh: 'JSON 输出 + 写 HTML 报告，给 CI 抓 exit code 同时人看。',
-                en: 'JSON output + HTML report, for CI exit code + human review.',
+                zh: 'JSON 输出 + 静默 gate，给 CI 抓 exit code 同时人看。',
+                en: 'JSON output + silent gate, for CI exit code + human review.',
             }),
             command: '<%= config.bin %> doctor --json --gate',
         },
@@ -79,12 +85,6 @@ export default class Doctor extends BaseCommand {
                 en: 'LLM model name, default sonnet.',
             }),
         }),
-        samples: Flags.string({
-            description: bilingual({
-                zh: '用例文件路径（.json/.yaml）。不传则按 target / cwd 顺序自动发现。',
-                en: 'Samples file path (.json/.yaml). Auto-detects from target / cwd if omitted.',
-            }),
-        }),
         timeout: Flags.string({
             description: bilingual({
                 zh: '单次 LLM 会话超时秒数，默认 600(10 分钟）。',
@@ -110,13 +110,6 @@ export default class Doctor extends BaseCommand {
                 en: 'Custom dimensions config file (YAML), appended after builtin 7. Each is either promptSection (LLM audit) or endpoint (POST skill snapshot to your service). Note: endpoint sends the full SKILL.md + sub-files to that URL — only enable for trusted configs/URLs.',
             }),
         }),
-        'static-only': Flags.boolean({
-            description: bilingual({
-                zh: '离线静态模式，只跑 4 条静态 rule(skill_readable / skill_metadata / dependencies_present / samples_contract_aligned），不调 LLM。',
-                en: 'Offline static mode: only 4 static rules, no LLM call.',
-            }),
-            default: false,
-        }),
         fix: Flags.boolean({
             description: bilingual({
                 zh: '交互式修复：根据 doctor 报告问题，用 LLM agent 修复 skill。',
@@ -129,6 +122,28 @@ export default class Doctor extends BaseCommand {
                 zh: 'LLM 推理 effort：low / medium / high / xhigh / max。',
                 en: 'LLM reasoning effort: low / medium / high / xhigh / max.',
             }),
+            parse: enumStringParser('--effort', ['low', 'medium', 'high', 'xhigh', 'max']),
+        }),
+        repeat: Flags.string({
+            description: bilingual({
+                zh: '健康度体检重复采样次数（self-consistency）。默认 2：并行跑 2 遍、finding 取并集并用 LLM 聚类归并同根因、标注支持度 k/N，压低单次采样方差。设 1 = 单次快速体检（不采样、不归并，最省）。',
+                en: 'Health-check repeat count (self-consistency). Default 2: runs 2 passes in parallel, unions findings, merges same root cause via an LLM pass, tags k/N support. Set 1 for a single quick pass (no sampling/merge, cheapest).',
+            }),
+            parse: integerStringParser('--repeat', { min: 1, max: 10 }),
+        }),
+        concurrency: Flags.string({
+            description: bilingual({
+                zh: '多次采样的并发数。默认 = --repeat（全并行，各遍相互独立，压墙钟时间）。设 1 = 串行。成本不变，只抬高瞬时并发（rate-limit 敏感时调小）。',
+                en: 'Concurrency across the repeated passes. Default = --repeat (full parallel; passes are independent, cuts wall-clock). Set 1 for serial. Cost unchanged; only raises peak concurrency (lower it if rate-limited).',
+            }),
+            parse: integerStringParser('--concurrency', { min: 1, max: 10 }),
+        }),
+        'static-only': Flags.boolean({
+            description: bilingual({
+                zh: '只跑静态检测（不调 LLM、不读 samples.json）：skill 可读性 / frontmatter 合法性 / 正文引用的脚本·CLI·文件·env 是否存在。CI 无 LLM 凭证或断网时用。',
+                en: 'Static checks only (no LLM, no samples.json): readability / frontmatter / existence of scripts·CLI·files·env referenced in the skill body. For CI without LLM creds / offline.',
+            }),
+            default: false,
         }),
     };
     async run() {
@@ -138,39 +153,21 @@ export default class Doctor extends BaseCommand {
             const target = args.target ?? null;
             const executorName = flags.executor ?? 'claude';
             const model = flags.model ?? 'sonnet';
-            // 默认 LLM 健康度审计(7 内置维度 + 用户注册的自定义维度);--static-only 切到
-            // 离线静态模式:只跑 4 条静态 rule,不调 LLM。CI 节点没装 claude/codex、本地断网
-            // 调试等场景。
-            const staticOnly = flags['static-only'];
-            const runHealthCheck = !staticOnly;
+            // omk doctor 默认 = 静态规则 + LLM 健康度审计(7 内置维度 + 用户注册的自定义维度);
+            // --static-only = 只跑静态检测(readable / metadata / 正文依赖,不调 LLM、不读 samples)。
+            // samples_contract_aligned 仍只归 eval preflight(它要 samples.json,与离线解耦)。
+            // 健康度体检重复采样次数(CLI flag --repeat → 内部 healthSamples 字段):默认 2,
+            // 设 1 = 单次快检。合法范围由 oclif parser 拦截。
+            const healthSamples = flags.repeat != null ? Number(flags.repeat) : 2;
+            // 归并策略:CLI 恒用 llm(硬逻辑,不暴露开关);samples=1 时 composer 自动跳过归并。
+            // 失败回退 string 仍在 composer 内兜底。programmatic runDoctor 默认仍是 string。
+            const healthMerge = 'llm';
+            // 并发数(--concurrency → healthConcurrency):默认不传(composer 取 = healthSamples 全并行);显式 ≥1 才覆盖。
+            const healthConcurrency = flags.concurrency != null ? Number(flags.concurrency) : undefined;
             const defaultTimeoutSec = 600;
             const timeoutSec = flags.timeout != null ? Number(flags.timeout) : defaultTimeoutSec;
             const timeoutMs = Math.max(1000, Math.floor((Number.isFinite(timeoutSec) ? timeoutSec : defaultTimeoutSec) * 1000));
             const cwd = process.cwd();
-            const samplesPath = flags.samples ? resolve(flags.samples) : findDoctorSamplesPath(target, cwd);
-            const deprecatedSamplesHint = !flags.samples && !samplesPath
-                ? findDoctorDeprecatedSamplesHint(target, cwd)
-                : null;
-            if (deprecatedSamplesHint) {
-                process.stderr.write(tCli('cli.common.deprecated_skill_samples_path', lang, {
-                    oldPath: deprecatedSamplesHint.oldPath,
-                    newPath: deprecatedSamplesHint.newPath,
-                }));
-            }
-            let samples;
-            let requires;
-            if (samplesPath) {
-                try {
-                    const { loadSamples } = await import('../../inputs/load-samples.js');
-                    const loaded = loadSamples(samplesPath);
-                    samples = loaded.samples;
-                    requires = loaded.requires;
-                }
-                catch (err) {
-                    const msg = err instanceof Error ? err.message : String(err);
-                    process.stderr.write(tCli('cli.common.warn_load_samples_failed', lang, { path: samplesPath, message: msg }));
-                }
-            }
             // 副作用 import: 注册 7 内置维度 spec + skill_health composer rule。
             await import('../../doctor/health/register.js');
             if (flags.dimensions) {
@@ -181,18 +178,25 @@ export default class Doctor extends BaseCommand {
                 }
             }
             const { runDoctor } = await import('../../doctor/index.js');
-            const { renderDoctorReportText, renderDoctorReportJson } = await import('../../doctor/renderer.js');
+            const { renderDoctorReportText, renderDoctorReportJson, renderDoctorActionPlanText } = await import('../../doctor/renderer.js');
             const { getRegisteredRules } = await import('../../doctor/rules.js');
             const { isComposerRule } = await import('../../types/doctor.js');
-            // 在线检查(LLM health composer + endpoint 自定义维度)默认跑;--static-only
-            // 离线模式只跑纯静态的内置 rule(无网络 / LLM)。endpoint rule 标了 external=true。
+            // 默认:静态规则 + 在线检查(LLM health composer + endpoint 自定义维度 external=true)。
+            // --static-only:只跑静态检测(纯静态内置 rule),但排除 samples_contract_aligned
+            // (那条要 samples.json,与离线解耦) → 且不加载 samples,依赖检查只扫 skill 正文。
+            const staticOnly = flags['static-only'];
             const isOnline = (r) => isComposerRule(r) || r.external === true;
             const rulesOverride = staticOnly
-                ? getRegisteredRules().filter((r) => !isOnline(r))
-                : getRegisteredRules().filter(isOnline);
+                ? getRegisteredRules().filter((r) => !isOnline(r) && r.id !== 'samples_contract_aligned')
+                : getRegisteredRules().filter((r) => r.id !== 'samples_contract_aligned');
             // 批量体检进度(per-skill,写 stderr)。--gate 是静默模式,不报进度;
             // --json 进度走 stderr 不污染 stdout 的 JSON。
             const onProgress = flags.gate ? undefined : makeDoctorProgress(lang);
+            // --effort 合法性由 oclif parser 拦截。这里仅做窄化后透传给健康审计。
+            const validEfforts = ['low', 'medium', 'high', 'xhigh', 'max'];
+            const effort = flags.effort && validEfforts.includes(flags.effort)
+                ? flags.effort
+                : undefined;
             let report;
             try {
                 report = await runDoctor({
@@ -202,10 +206,12 @@ export default class Doctor extends BaseCommand {
                     model,
                     timeoutMs,
                     lang,
-                    runHealthCheck,
+                    runHealthCheck: !staticOnly,
+                    healthSamples,
+                    healthMerge,
+                    healthConcurrency,
+                    effort,
                     rules: rulesOverride,
-                    samples,
-                    requires,
                     onProgress,
                 });
             }
@@ -230,12 +236,10 @@ export default class Doctor extends BaseCommand {
                         ? `doctor failed: ${report.totals.fail} 个 skill 未通过 (${report.totals.warn} warn / ${report.totals.pass} pass)`
                         : `doctor failed: ${report.totals.fail} skills did not pass (${report.totals.warn} warn / ${report.totals.pass} pass)`;
                     console.error(summary);
+                    process.stderr.write(renderDoctorActionPlanText(report, lang));
                 }
             }
             else {
-                if (samplesPath && samples) {
-                    process.stderr.write(tCli('cli.doctor.samples_detected', lang, { path: samplesPath }) + '\n');
-                }
                 renderDoctorReportText(report, lang);
             }
             persistDoctorReport(report, flags['output-dir']
@@ -248,7 +252,7 @@ export default class Doctor extends BaseCommand {
                     throw new CliExit(0);
                 }
                 const { runDoctorFix } = await import('../../doctor/fixer.js');
-                const changed = await runDoctorFix({ report: existing, executorName, model, timeoutMs });
+                const changed = await runDoctorFix({ report: existing, executorName, model, timeoutMs, effort });
                 throw new CliExit(changed ? 0 : (existing.outcome === 'failed' ? 1 : 0));
             }
             throw new CliExit(report.outcome === 'failed' ? 1 : 0);

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

@@ -152,6 +152,14 @@ async function emitBatchVerdict(report, reportsDir, values, lang) {
     for (const result of results) {
         emitVerdictText(`  ${result.verdict.level}: ${result.treatment} — ${result.verdict.headline}`);
     }
+    const next = lang === 'zh'
+        ? (failed === 0
+            ? '全部通过，可以进入发布流程：请留存这些报告作为发布证据；如果是受管 skill，再运行 `omk promote`。'
+            : '先处理未通过的 skill；逐个打开对应报告，看最差层和失败用例后再重跑。')
+        : (failed === 0
+            ? 'all skills passed and are ready for release: keep these reports as release evidence; for managed skills, run `omk promote`.'
+            : 'fix the failing skills first; open each report, inspect the weakest layer and failing samples, then re-run.');
+    emitVerdictText(tCli('cli.run.batch_verdict_next_step', lang, { next }));
     return failed === 0 ? 0 : 1;
 }
 async function announceSavedReport({ report, filePath, reportsDir, values, lang, }) {

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

@@ -58,11 +58,12 @@ export interface DoctorFlags {
     gate: boolean;
     executor?: string;
     model?: string;
-    samples?: string;
     timeout?: string;
     'output-dir'?: string;
     dimensions?: string;
-    'static-only': boolean;
+    repeat?: string;
+    concurrency?: string;
+    effort?: string;
 }
 export type EvalArgs = Record<string, never>;
 export interface EvalFlags {

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.deprecated_skill_samples_path' | 'cli.common.samples_not_found' | '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.samples_detected' | '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.report_not_found' | 'cli.common.no_judge_model' | 'cli.common.judge_models_single_only' | 'cli.common.warn_load_samples_failed' | 'cli.common.deprecated_skill_samples_path' | 'cli.common.samples_not_found' | '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

@@ -91,10 +91,6 @@ export const commonDict = {
         zh: '未在 {path} 下发现 skill 文件。\n  doctor 期望 .md 文件、目录(包含 .md 或 SKILL.md)或 cwd 下的 skills/ 子目录。',
         en: 'No skills found at {path}.\n  doctor expects a .md file, a directory (containing .md or SKILL.md), or skills/ under cwd.',
     },
-    'cli.doctor.samples_detected': {
-        zh: '✓ 使用评测用例文件：{path}',
-        en: '✓ Using eval samples file: {path}',
-    },
     'cli.doctor.progress_skill_start': {
         zh: '{prefix}{skill} ⏳ 体检中...\n',
         en: '{prefix}{skill} ⏳ checking...\n',

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_child_report_missing' | 'cli.run.eval_complete' | 'cli.run.tally' | 'cli.run.report_saved' | 'cli.run.evidence_recorded' | '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.skip_connectivity_warning';
+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_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.skip_connectivity_warning';
 export declare const runDict: Record<RunMessageKey, CliMessage>;

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

@@ -87,6 +87,10 @@ export const runDict = {
         zh: '批量评测结论：{status}（{passed}/{total} 通过）',
         en: 'Batch verdict: {status} ({passed}/{total} passed)',
     },
+    'cli.run.batch_verdict_next_step': {
+        zh: '  下一步：{next}',
+        en: '  Next: {next}',
+    },
     'cli.run.batch_child_report_missing': {
         zh: '⚠ 子报告缺失：{id}，将按不可 ship 处理。\n',
         en: '⚠ Child report missing: {id}; treating it as not shippable.\n',

package/dist/doctor/endpoint-rule.js

@@ -222,7 +222,7 @@ export function makeEndpointRule(spec, fetchFn = fetch) {
         severity: spec.severity,
         // 自定义 key 不在 DOCTOR_MESSAGES 字典里,renderer 会 fallback 到 ruleId(=spec.id)。
         labelKey: `cli.doctor.endpoint.${spec.id}`,
-        // 网络检查,与 health composer 同档:默认跑,--static-only 跳过。
+        // 网络检查,与 health composer 同档:CLI doctor 默认跑(在线检查)。
         external: true,
         async check(ctx) {
             // SSRF 防护(defense-in-depth):endpoint 会收到 skill 完整快照,且响应