oh-my-knowledge 0.35.0 → 0.37.0

package/dist/observability/soft-standards/llm-extractor.js

@@ -2,7 +2,7 @@ import { mkdirSync, writeFileSync } from 'node:fs';
 import { createExecutor } from '../../executors/index.js';
 import { readPromptDocument } from '../../shared/llm-prompts/index.js';
 import { DEFAULT_LLM_ENHANCED_REVIEW_MODEL, PROMPTS_DIR, SOFT_STANDARD_PROMPT_ID, SOFT_STANDARD_PROMPT_VERSION, } from './constants.js';
-import { hashText, loadExisting, markStale, skillDerivedStandardsDir, skillDerivedStandardsPath, } from './skill-standards-store.js';
+import { hashText, loadExisting, markStale, SKILL_DERIVED_STANDARDS_SCHEMA_VERSION, skillDerivedStandardsDir, skillDerivedStandardsPath, } from './skill-standards-store.js';
 import { evaluateRuntimeStandardNodes } from './runtime-evaluator.js';
 export async function extractSkillSoftStandards(options) {
     const { observationsDir, skillChain } = options;
@@ -47,8 +47,8 @@ export async function extractSkillSoftStandards(options) {
         source: 'llm_soft_standard',
     }));
     const record = {
-        reportKind: 'observe-skill-derived-standards',
-        schemaVersion: 1,
+        kind: 'observe-skill-derived-standards',
+        schemaVersion: SKILL_DERIVED_STANDARDS_SCHEMA_VERSION,
         skillName: skillChain.skillName,
         sourceSkillPath: skillChain.definition.path,
         sourceHash,

package/dist/observability/soft-standards/skill-standards-store.d.ts

@@ -1,4 +1,5 @@
 import type { ResolveSkillStandardsOptions, ResolvedSkillStandards, SkillDerivedStandardStatus, SkillDerivedStandards } from './types.js';
+export declare const SKILL_DERIVED_STANDARDS_SCHEMA_VERSION = 2;
 export declare function skillDerivedStandardsDir(observationsDir: string): string;
 export declare function skillDerivedStandardsPath(observationsDir: string, skillName: string): string;
 export declare function loadSkillDerivedStandards(observationsDir: string): Record<string, SkillDerivedStandards>;

package/dist/observability/soft-standards/skill-standards-store.js

@@ -2,6 +2,7 @@ import { createHash } from 'node:crypto';
 import { existsSync, mkdirSync, readFileSync, readdirSync, writeFileSync } from 'node:fs';
 import { join } from 'node:path';
 import { buildObservationSkillChain } from '../skill-chain.js';
+export const SKILL_DERIVED_STANDARDS_SCHEMA_VERSION = 2;
 export function skillDerivedStandardsDir(observationsDir) {
     return join(observationsDir, 'skill-derived');
 }
@@ -17,8 +18,8 @@ export function loadSkillDerivedStandards(observationsDir) {
         if (!file.endsWith('.json'))
             continue;
         try {
-            const parsed = JSON.parse(readFileSync(join(dir, file), 'utf-8'));
-            if (isSkillDerivedStandards(parsed))
+            const parsed = normalizeSkillDerivedStandards(JSON.parse(readFileSync(join(dir, file), 'utf-8')));
+            if (parsed)
                 out[parsed.skillName] = parsed;
         }
         catch {
@@ -51,11 +52,13 @@ export function updateSkillDerivedStandardStatus(observationsDir, skillName, sta
 }
 export function resolveSkillStandards(skillName, options) {
     const skillChain = options.skillChain ?? buildObservationSkillChain(skillName, options.cwd ?? process.cwd());
-    const derived = options.derivedStandards
-        ? isSkillDerivedStandards(options.derivedStandards)
-            ? options.derivedStandards
-            : options.derivedStandards[skillName]
-        : loadSkillDerivedStandards(options.observationsDir)[skillName];
+    const directDerived = normalizeSkillDerivedStandards(options.derivedStandards);
+    const mappedDerived = options.derivedStandards && !isSkillDerivedStandards(options.derivedStandards)
+        ? normalizeSkillDerivedStandards(options.derivedStandards[skillName])
+        : undefined;
+    const derived = directDerived
+        ?? mappedDerived
+        ?? loadSkillDerivedStandards(options.observationsDir)[skillName];
     const active = [];
     const candidates = [];
     const hasFrontmatterHardRules = skillChain.healthCheck.hardRules.declared && skillChain.healthCheck.hardRules.rules.length > 0;
@@ -112,8 +115,7 @@ export function loadExisting(path) {
     if (!existsSync(path))
         return undefined;
     try {
-        const parsed = JSON.parse(readFileSync(path, 'utf-8'));
-        return isSkillDerivedStandards(parsed) ? parsed : undefined;
+        return normalizeSkillDerivedStandards(JSON.parse(readFileSync(path, 'utf-8')));
     }
     catch {
         return undefined;
@@ -129,13 +131,74 @@ export function markStale(record, generatedAt) {
     };
 }
 export function isSkillDerivedStandards(value) {
+    return normalizeSkillDerivedStandards(value) !== undefined;
+}
+function normalizeSkillDerivedStandards(value) {
     if (!value || typeof value !== 'object')
-        return false;
+        return undefined;
     const item = value;
-    return item.reportKind === 'observe-skill-derived-standards'
-        && item.schemaVersion === 1
-        && typeof item.skillName === 'string'
-        && Array.isArray(item.standards);
+    const kind = item.kind === 'observe-skill-derived-standards' ? item.kind : null;
+    if (!kind)
+        return undefined;
+    if (item.schemaVersion !== SKILL_DERIVED_STANDARDS_SCHEMA_VERSION)
+        return undefined;
+    if (typeof item.skillName !== 'string' || typeof item.generatedAt !== 'string' || typeof item.model !== 'string' || typeof item.executor !== 'string') {
+        return undefined;
+    }
+    if (typeof item.promptId !== 'string' || typeof item.promptVersion !== 'string' || !Array.isArray(item.standards))
+        return undefined;
+    const standards = item.standards
+        .map(normalizeSkillDerivedStandard)
+        .filter((record) => record !== null);
+    return {
+        kind: 'observe-skill-derived-standards',
+        schemaVersion: SKILL_DERIVED_STANDARDS_SCHEMA_VERSION,
+        skillName: item.skillName,
+        ...(typeof item.sourceSkillPath === 'string' ? { sourceSkillPath: item.sourceSkillPath } : {}),
+        ...(typeof item.sourceHash === 'string' ? { sourceHash: item.sourceHash } : {}),
+        generatedAt: item.generatedAt,
+        model: item.model,
+        executor: item.executor,
+        promptId: item.promptId,
+        promptVersion: item.promptVersion,
+        ...(typeof item.promptHash === 'string' ? { promptHash: item.promptHash } : {}),
+        ...(typeof item.runtimeEvidenceHash === 'string' ? { runtimeEvidenceHash: item.runtimeEvidenceHash } : {}),
+        ...(item.enhancedReview && typeof item.enhancedReview === 'object' ? { enhancedReview: item.enhancedReview } : {}),
+        standards,
+    };
+}
+function normalizeSkillDerivedStandard(value) {
+    if (!value || typeof value !== 'object')
+        return null;
+    const item = value;
+    const standardKind = item.standardKind === 'hard_rule_candidate' || item.standardKind === 'workflow_candidate'
+        ? item.standardKind
+        : item.kind === 'hard_rule_candidate' || item.kind === 'workflow_candidate'
+            ? item.kind
+            : null;
+    if (!standardKind)
+        return null;
+    if (typeof item.id !== 'string' || typeof item.title !== 'string' || typeof item.body !== 'string')
+        return null;
+    if (item.status !== 'pending_review' && item.status !== 'author_confirmed' && item.status !== 'rejected' && item.status !== 'stale') {
+        return null;
+    }
+    if (item.source !== 'llm_soft_standard')
+        return null;
+    if (item.confidence !== 'low' && item.confidence !== 'medium' && item.confidence !== 'high')
+        return null;
+    if (!Array.isArray(item.evidence) || item.evidence.some((entry) => typeof entry !== 'string'))
+        return null;
+    return {
+        id: item.id,
+        standardKind,
+        status: item.status,
+        title: item.title,
+        body: item.body,
+        source: item.source,
+        confidence: item.confidence,
+        evidence: item.evidence,
+    };
 }
 function candidateRank(status) {
     if (status === 'pending_review')

package/dist/observability/soft-standards/types.d.ts

@@ -14,8 +14,8 @@ export interface SkillDerivedStandard {
     evidence: string[];
 }
 export interface SkillDerivedStandards {
-    reportKind: 'observe-skill-derived-standards';
-    schemaVersion: 1;
+    kind: 'observe-skill-derived-standards';
+    schemaVersion: 2;
     skillName: string;
     sourceSkillPath?: string;
     sourceHash?: string;

package/dist/renderer/doctor-detail-renderer.d.ts

@@ -0,0 +1,9 @@
+/**
+ * Doctor 报告详情页 — 逐条规则的健康检查结果。
+ *
+ * 主次：Hero 综合健康分 → 统计条（通过/警告/失败）→ 规则列表（失败/警告高亮在前，
+ * 通过项折叠）。每条规则展开显示 finding 描述 + 修复建议。
+ */
+import { type SkillReportContext } from './report-shell.js';
+import type { Lang, DoctorReport } from '../types/index.js';
+export declare function renderDoctorDetail(report: DoctorReport, skillName: string, langQ: string, lang?: Lang, skillContext?: SkillReportContext): string;

package/dist/renderer/doctor-detail-renderer.js

@@ -0,0 +1,114 @@
+/**
+ * Doctor 报告详情页 — 逐条规则的健康检查结果。
+ *
+ * 主次：Hero 综合健康分 → 统计条（通过/警告/失败）→ 规则列表（失败/警告高亮在前，
+ * 通过项折叠）。每条规则展开显示 finding 描述 + 修复建议。
+ */
+import { reportShell, healthColor } from './report-shell.js';
+import { icon as svgIcon } from './icons.js';
+import { e, DEFAULT_LANG } from './layout.js';
+function getDetail(r) {
+    return (r.detail ?? {});
+}
+const SORT_ORDER = { fail: 0, warn: 1, pass: 2, skipped: 3 };
+function fmtTs(ts) {
+    try {
+        const d = new Date(ts);
+        return `${d.getFullYear()}-${String(d.getMonth() + 1).padStart(2, '0')}-${String(d.getDate()).padStart(2, '0')} ${String(d.getHours()).padStart(2, '0')}:${String(d.getMinutes()).padStart(2, '0')}`;
+    }
+    catch {
+        return ts;
+    }
+}
+function renderRuleCard(r, lang, forceOpen = false) {
+    const det = getDetail(r);
+    const name = det.displayName ?? r.ruleId;
+    const findings = det.findings ?? [];
+    const errors = findings.filter((f) => f.level === '错误');
+    const warns = findings.filter((f) => f.level === '警告');
+    const tips = findings.filter((f) => f.level === '建议');
+    const cardCls = r.status === 'fail' ? 'fail' : r.status === 'warn' ? 'warn' : 'pass';
+    const dotCls = r.status === 'fail' ? 'fail' : r.status === 'warn' ? 'warn' : r.status === 'skipped' ? 'skip' : 'pass';
+    const badges = [];
+    if (errors.length)
+        badges.push(`<span class="rs-badge rs-badge--fail">${errors.length} ${lang === 'zh' ? '错误' : 'err'}</span>`);
+    if (warns.length)
+        badges.push(`<span class="rs-badge rs-badge--warn">${warns.length} ${lang === 'zh' ? '警告' : 'warn'}</span>`);
+    if (!errors.length && !warns.length)
+        badges.push(`<span class="rs-badge rs-badge--pass">${lang === 'zh' ? '通过' : 'pass'}</span>`);
+    const head = `<span class="rs-card-dot rs-card-dot--${dotCls}"></span><span class="rs-card-name">${e(name)}</span><span class="rs-card-badges">${badges.join('')}</span>`;
+    const allFindings = [...errors, ...warns, ...tips];
+    if (allFindings.length === 0) {
+        return `<div class="rs-card rs-card--${cardCls}"><div class="rs-card-head" style="cursor:default">${head}</div></div>`;
+    }
+    const findingHtml = allFindings.map((f) => {
+        const isErr = f.level === '错误';
+        const mIcon = isErr
+            ? svgIcon('x', { size: 14, style: 'color:#dc2626;vertical-align:-2px;margin-right:5px' })
+            : svgIcon('warn', { size: 14, style: `color:${f.level === '警告' ? '#d97706' : '#637083'};vertical-align:-2px;margin-right:5px` });
+        return `<div class="rs-finding rs-finding--${isErr ? 'err' : 'warn'}">
+      <div class="rs-finding-desc">${mIcon}${e(f.description ?? '')}</div>
+      ${f.suggestion ? `<div class="rs-finding-sug">${svgIcon('bulb', { size: 13, style: 'vertical-align:-2px;margin-right:5px' })}${e(f.suggestion)}</div>` : ''}
+    </div>`;
+    }).join('');
+    return `<details class="rs-card rs-card--${cardCls}" ${errors.length || forceOpen ? 'open' : ''}>
+    <summary class="rs-card-head">${head}</summary>
+    <div class="rs-card-body">${findingHtml}</div>
+  </details>`;
+}
+function renderSkillBlock(sk, lang) {
+    const rules = sk.results.filter((r) => !r.ruleId.endsWith(':_summary'));
+    const pass = rules.filter((r) => r.status === 'pass').length;
+    const warn = rules.filter((r) => r.status === 'warn').length;
+    const fail = rules.filter((r) => r.status === 'fail').length;
+    const total = rules.length;
+    const sorted = [...rules].sort((a, b) => (SORT_ORDER[a.status] ?? 9) - (SORT_ORDER[b.status] ?? 9));
+    const issues = sorted.filter((r) => r.status === 'fail' || r.status === 'warn');
+    const oks = sorted.filter((r) => r.status === 'pass' || r.status === 'skipped');
+    const zh = lang === 'zh';
+    const stats = `<div class="rs-stats">
+    <div class="rs-stat rs-stat--pass"><span class="rs-stat-num">${pass}</span><span class="rs-stat-label">${zh ? '通过' : 'Pass'}</span></div>
+    <div class="rs-stat rs-stat--warn"><span class="rs-stat-num">${warn}</span><span class="rs-stat-label">${zh ? '警告' : 'Warn'}</span></div>
+    <div class="rs-stat rs-stat--fail"><span class="rs-stat-num">${fail}</span><span class="rs-stat-label">${zh ? '失败' : 'Fail'}</span></div>
+    <div class="rs-stat rs-stat--total"><span class="rs-stat-num">${total}</span><span class="rs-stat-label">${zh ? '规则总数' : 'Rules'}</span></div>
+  </div>`;
+    const issuesHtml = issues.length
+        ? `<div class="rs-panel"><div class="rs-panel-title">${zh ? '待处理项' : 'Issues'} <small>${issues.length}</small></div>${issues.map((r, i) => renderRuleCard(r, lang, i === 0)).join('')}</div>`
+        : `<div class="rs-panel"><div class="rs-empty">${zh ? '✓ 所有规则通过，无待处理项' : '✓ All rules passed'}</div></div>`;
+    const oksHtml = oks.length
+        ? `<div class="rs-panel"><details class="rs-fold"><summary>${zh ? `展开 ${oks.length} 条通过的规则` : `Show ${oks.length} passed rules`}</summary>${oks.map((r) => renderRuleCard(r, lang)).join('')}</details></div>`
+        : '';
+    return stats + issuesHtml + oksHtml;
+}
+export function renderDoctorDetail(report, skillName, langQ, lang = DEFAULT_LANG, skillContext) {
+    const zh = lang === 'zh';
+    // 优先展示点进来的 skill；找不到就展示全部
+    const targetSkill = report.skills.find((s) => s.skillName === skillName) ?? report.skills[0];
+    const rules = (targetSkill?.results ?? []).filter((r) => !r.ruleId.endsWith(':_summary'));
+    const pass = rules.filter((r) => r.status === 'pass').length;
+    const warn = rules.filter((r) => r.status === 'warn').length;
+    const total = rules.length;
+    const score = total > 0 ? Math.round(((pass + warn * 0.5) / total) * 100) : null;
+    // 与 eval「综合分 / 5」统一成「X分 / 满分」形式;圆心显原生分数,等级由颜色传达。
+    const scoreLabel = zh ? '健康分 / 100' : 'Health / 100';
+    const body = targetSkill
+        ? renderSkillBlock(targetSkill, lang)
+        : `<div class="rs-panel"><div class="rs-empty">${zh ? '报告无 skill 数据' : 'No skill data'}</div></div>`;
+    return reportShell({
+        dim: 'doctor',
+        icon: '🩺',
+        kindTitle: zh ? '健康体检报告' : 'Doctor Report',
+        skillName: targetSkill?.skillName ?? skillName,
+        metaItems: [`${svgIcon('clock', { size: 13 })} ${e(fmtTs(report.timestamp))}`, `${svgIcon('chip', { size: 13 })} ${e(report.model)}`, e(report.id)],
+        // Hero 展示「综合健康」总分(跨维度,与首页同口径),切 tab 不变;本次体检明细(规则/计数)在 body。
+        // 无 skillContext 时回退显本次体检健康分。
+        score: skillContext ? skillContext.overall.score : score,
+        scoreText: skillContext && skillContext.overall.score != null ? String(skillContext.overall.score) : undefined,
+        ringColor: skillContext ? healthColor(skillContext.overall.score) : undefined,
+        scoreLabel: skillContext ? (zh ? '综合健康' : 'Health') : scoreLabel,
+        backHref: `/${langQ}`,
+        backLabel: zh ? '返回列表' : 'Back to list',
+        body,
+        skillContext,
+    }, lang);
+}

package/dist/renderer/html-renderer.d.ts

@@ -1,9 +1,10 @@
 /**
  * HTML report renderer — orchestrates sub-modules.
  */
+import { type SkillReportContext } from './report-shell.js';
 import type { BatchEvaluationReport, EvaluationReport, Report, ReportDocument, Lang } from '../types/index.js';
 export declare function renderRunList(runs: ReportDocument[], lang?: Lang): string;
-export declare function renderRunDetail(report: EvaluationReport | null, lang?: Lang): string;
+export declare function renderRunDetail(report: EvaluationReport | null, lang?: Lang, skillContext?: SkillReportContext): string;
 export declare function renderBatchEvaluationDetail(report: BatchEvaluationReport | null, lang?: Lang): string;
-export declare function renderReportDocumentDetail(report: ReportDocument | null, lang?: Lang): string;
+export declare function renderReportDocumentDetail(report: ReportDocument | null, lang?: Lang, skillContext?: SkillReportContext): string;
 export declare function renderTrendsPage(variantName: string, runs: Report[], lang?: Lang): string;

package/dist/renderer/html-renderer.js

@@ -3,9 +3,10 @@
  */
 import { e, fmtCost, fmtDuration, fmtKnownCost, COLORS, DEFAULT_LANG, t, layout } from './layout.js';
 import { renderAgentOverview, renderAnalysis, renderKnowledgeInteractionSection, renderMethodologyAudit, renderScoringModal, renderSummaryCards, renderVarianceComparisons, renderVerdictPill, levelLabel, levelTooltip, } from './summary.js';
-import { renderSampleTable } from './table.js';
 import { renderTestView, TEST_VIEW_CSS, TEST_VIEW_JS } from './test-view.js';
 import { renderTrendsBody } from './trends.js';
+import { reportShell, healthColor } from './report-shell.js';
+import { icon } from './icons.js';
 import { computeVerdict } from '../eval-core/verdict.js';
 // v0.21 B.4 — 列表页 status pill 用的 dot. PROGRESS/REGRESS 实心(强信号),
 // CAUTIOUS 三角(警示),NOISE 空心圆(有信号但无效果),UNDERPOWERED 部分填充
@@ -21,7 +22,7 @@ function levelDot(level) {
     }
 }
 function isEvaluationReport(document) {
-    return document.reportKind === 'evaluation';
+    return document.kind === 'evaluation';
 }
 function scoreOf(summary) {
     return summary?.avgCompositeScore ?? summary?.avgLlmScore ?? null;
@@ -163,7 +164,7 @@ export function renderRunList(runs, lang = DEFAULT_LANG) {
             `<span class="rc-score-num rc-score-num--${cls}">${score.toFixed(2)}</span></div>`;
     };
     const cards = runs.map((run) => {
-        if (run.reportKind === 'batch-evaluation') {
+        if (run.kind === 'batch-evaluation') {
             const m = run.meta;
             const scores = run.items.length > 0
                 ? run.items.map((item) => {
@@ -316,7 +317,7 @@ export function renderRunList(runs, lang = DEFAULT_LANG) {
     </main>
   `, lang);
 }
-export function renderRunDetail(report, lang = DEFAULT_LANG) {
+export function renderRunDetail(report, lang = DEFAULT_LANG, skillContext) {
     const langQ = lang === DEFAULT_LANG ? '' : `?lang=${lang}`;
     if (!report) {
         return layout(t('title', lang), `
@@ -329,11 +330,24 @@ export function renderRunDetail(report, lang = DEFAULT_LANG) {
     const m = report.meta;
     const variants = m.variants || [];
     const summary = report.summary || {};
-    const results = report.results || [];
     const cards = renderSummaryCards(variants, summary, lang, report.variance);
     const methodologyAudit = renderMethodologyAudit(report, variants, summary, lang);
-    const verdictPill = renderVerdictPill(report, lang);
-    const sampleTable = renderSampleTable(variants, results, lang);
+    // 单 variant(SOLO)不显示 verdict — 对照才有的概念,单组是噪音/误导(见融合设计)。
+    const verdictPill = variants.length > 1 ? renderVerdictPill(report, lang) : '';
+    // Hero：综合分 ring（取 treatment = 最后一个 variant 的综合分）+ skill 名 + 时间戳。
+    const treatmentName = variants[variants.length - 1];
+    const heroComposite = treatmentName ? (summary[treatmentName]?.avgCompositeScore ?? null) : null;
+    const heroScore100 = heroComposite != null ? Math.round((heroComposite / 5) * 100) : null;
+    const heroSkillName = report.meta.evolve?.skillName ?? treatmentName ?? report.id;
+    const heroTs = (() => {
+        try {
+            const d = new Date(report.meta.timestamp);
+            return `${d.getFullYear()}-${String(d.getMonth() + 1).padStart(2, '0')}-${String(d.getDate()).padStart(2, '0')} ${String(d.getHours()).padStart(2, '0')}:${String(d.getMinutes()).padStart(2, '0')}`;
+        }
+        catch {
+            return '';
+        }
+    })();
     const totalExecCost = Object.values(summary).reduce((s, v) => s + (v.totalExecCostUSD || 0), 0);
     // 任一 variant exec cost 未报告 → 总 cost 不可靠,renderer 显示「—」+ tooltip。
     // 全部 reported(undefined / true)才正常显示 USD 数字。
@@ -420,14 +434,6 @@ export function renderRunDetail(report, lang = DEFAULT_LANG) {
   ` : '';
     // v0.30 重设计 — 评分/单测各自独立 title + meta strip(不共享头部)。
     // 顶部只保留极简身份带:返回链接 + run-id pill。完整 H1 + meta tags 移进各自 tab 面板。
-    const idStamp = (() => {
-        const idMatch = /^.+?-(\d{8})-(\d{4})$/.exec(report.id);
-        if (idMatch) {
-            const [, ymd, hm] = idMatch;
-            return `${ymd.slice(0, 4)}-${ymd.slice(4, 6)}-${ymd.slice(6, 8)} ${hm.slice(0, 2)}:${hm.slice(2, 4)}`;
-        }
-        return '';
-    })();
     // ──────────── 评分 tab 的 meta tags(原版全部留这里) ────────────
     const scoreMetaTags = `<div class="meta-tags">
     <span class="meta-tag">${t('model', lang)}: ${e(m.model)}</span>
@@ -469,71 +475,61 @@ export function renderRunDetail(report, lang = DEFAULT_LANG) {
         ${Object.entries(m.blindMap || {}).map(([label, real]) => `<div style="font-size:13px;color:var(--text-secondary)"><strong>Variant ${e(label)}</strong> → ${e(real)}</div>`).join('')}
       </div>
     </div>` : '';
-    // ──────────── 单测 tab 的 meta tags(只有 sample 维度数据,不含 cost / model) ────────────
-    const sampleSnaps = report.sampleSnapshots ?? {};
-    let testPassed = 0;
-    let testFailed = 0;
-    let testTripwire = 0;
-    for (const r of (report.results || [])) {
-        const v = m.variants?.[0] ? r.variants[m.variants[0]] : null;
-        if (!v)
-            continue;
-        const passed = (v.assertions?.details ?? []).every((d) => d.passed);
-        if (passed)
-            testPassed += 1;
-        else if (sampleSnaps[r.sample_id]?.tripwire || (v.diagnostic?.rootCause ?? []).includes('tripwire_intentional'))
-            testTripwire += 1;
-        else
-            testFailed += 1;
-    }
-    const testMetaTags = `<div class="meta-tags">
-    <span class="meta-tag">${m.sampleCount} ${lang === 'zh' ? '用例' : 'samples'}</span>
-    <span class="meta-tag" style="color:var(--green)">✓ ${testPassed} ${lang === 'zh' ? '通过' : 'passed'}</span>
-    ${testFailed > 0 ? `<span class="meta-tag" style="color:var(--red)">✗ ${testFailed} ${lang === 'zh' ? '失败' : 'failed'}</span>` : ''}
-    ${testTripwire > 0 ? `<span class="meta-tag" style="color:#7a6b89">◆ ${testTripwire} ${lang === 'zh' ? '诱错触发' : 'tripwire'}</span>` : ''}
-    <span class="meta-tag">${(m.variants || []).join(' · ')}</span>
-  </div>`;
-    return layout(`${report.id}`, `
-    <main>
-    <nav class="nav"><a href="/${langQ}" data-i18n="backToList">${t('backToList', lang)}</a></nav>
-    <div class="report-id-line">
-      <code class="report-id-pill">${e(report.id)}</code>
-      ${idStamp ? `<span class="report-id-stamp">${e(idStamp)}</span>` : ''}
-    </div>
-
-    <div class="omk-view-tabs" role="tablist">
-      <button type="button" class="omk-view-tab omk-view-tab--active" data-view="score" onclick="omkSwitchView('score')">${lang === 'zh' ? '📊 评测视角' : '📊 Score view'}</button>
-      <button type="button" class="omk-view-tab" data-view="test" onclick="omkSwitchView('test')">${lang === 'zh' ? '✅ 功能视角' : '✅ Functional view'}</button>
-    </div>
-
-    <div class="omk-view-panel" data-view="score">
-    <p class="subtitle">${lang === 'zh' ? '主观质量打分(judge)+ 客观断言聚合,产出综合评分与 verdict' : 'Subjective judge + objective assertions aggregated into composite + verdict'}</p>
-    ${verdictPill}
-    ${scoreMetaTags}
+    // ──────────── 融合单页:Hero → 结论(verdict+六维) → 折叠次要区 → 统一逐用例 ────────────
+    const agentOverview = renderAgentOverview(variants, summary, lang);
+    const varianceSection = renderVarianceComparisons(report.variance, lang, Boolean(report.meta.layeredStats), summary);
+    const analysisSection = renderAnalysis(report, lang);
+    const knowledgeSection = renderKnowledgeInteractionSection(report.analysis?.coverage, report.analysis?.gapReports, lang);
+    // 折叠区 A:实验如何跑的(配置 / 运行参数 / Agent 概览 / 审计指纹)— 复现/审计才看。
+    const setupInner = [
+        scoreMetaTags,
+        (variantConfigSection || agentOverview) ? `<div class="ctx-strip">${variantConfigSection}${agentOverview}</div>` : '',
+        auditFingerprints,
+    ].filter(Boolean).join('');
+    const setupFold = setupInner
+        ? `<details class="ev-fold"><summary>${lang === 'zh' ? '实验配置与运行参数' : 'Experiment setup & run params'}</summary><div class="ev-fold-body">${setupInner}</div></details>`
+        : '';
+    // 折叠区 B:深入分析(方法学审计 / 方差 / 自动分析 / 知识交互)— 想验证测量学时展开。
+    const analysisInner = [methodologyAudit, varianceSection, analysisSection, knowledgeSection].filter(Boolean).join('');
+    const analysisFold = analysisInner
+        ? `<details class="ev-fold"><summary>${lang === 'zh' ? '方法学审计 · 方差 · 自动分析 · 知识交互' : 'Methodology audit · variance · analysis · knowledge'}</summary><div class="ev-fold-body">${analysisInner}</div></details>`
+        : '';
+    // 「结论」panel:多 variant = verdict 一句话领头 + 六维表;单 variant = 直接六维分布。
+    const conclusionPanel = `<section class="ev-conclusion">${verdictPill}${cards}</section>`;
+    const samplesSubtitle = lang === 'zh'
+        ? '每条用例既是评分卡也是功能测试：综合分 + 三层分布、断言、诊断与执行轨迹合一'
+        : 'Each sample is both a score card and a functional test: composite + layered scores, assertions, diagnostics, and trace combined';
+    const body = `
+    ${conclusionPanel}
     ${blindReveal}
-
-    <section>${cards}</section>
-    <div class="ctx-strip">${variantConfigSection}${renderAgentOverview(variants, summary, lang)}</div>
-    ${auditFingerprints}
-    ${methodologyAudit}
-
-    ${renderVarianceComparisons(report.variance, lang, Boolean(report.meta.layeredStats), summary)}
-
-    ${renderAnalysis(report, lang)}
-
-    ${renderKnowledgeInteractionSection(report.analysis?.coverage, report.analysis?.gapReports, lang)}
-
-    <section>${sampleTable}</section>
-    </div>
-
-    <div class="omk-view-panel" data-view="test" hidden>
-    <p class="subtitle">${lang === 'zh' ? '每条用例当一条功能测试看:期望(rubric/assertions/mocks)、实际执行(trace)、失败时给诊断建议' : 'Each sample as a functional test: expected(rubric/assertions/mocks), actual(trace), diagnostic suggestions on failure'}</p>
-    ${testMetaTags}
-    ${renderTestView(report, lang)}
-    </div>
-
-    <style>${TEST_VIEW_CSS}
-/* Context strip — 实验配置 + Agent 概览并排,作为辅助元数据,视觉权重低于综合分 hero / 六维 / 用例详情 */
+    ${setupFold}
+    ${analysisFold}
+    <section class="ev-samples">
+      <h2 class="ev-samples-h">${lang === 'zh' ? '逐用例' : 'Per-sample'}</h2>
+      <p class="subtitle ev-samples-sub">${samplesSubtitle}</p>
+      ${renderTestView(report, lang)}
+    </section>${TEST_VIEW_JS ? `<script>${TEST_VIEW_JS}</script>` : ''}`;
+    const evalCss = `${TEST_VIEW_CSS}
+/* 结论 panel — verdict(多 variant)+ 六维,白底卡片成「答案」模块,与 hero / fold 同一视觉语言 */
+.ev-conclusion{background:#fff;border:1px solid #e4e8f1;border-radius:12px;box-shadow:0 8px 24px rgba(31,41,55,4%);padding:18px 24px 20px;margin:0 0 16px}
+.ev-conclusion>:first-child{margin-top:0}
+.ev-conclusion>h2:first-child{margin-top:2px}
+/* 次要折叠区 — 实验配置 / 方法学 / 方差 / 分析,默认收起,审计/复现时展开 */
+.ev-fold{margin:10px 0;background:var(--bg-surface);border:1px solid var(--border);border-radius:var(--radius)}
+.ev-fold>summary{cursor:pointer;list-style:none;font-size:13px;font-weight:500;color:var(--text-secondary);padding:12px 16px}
+.ev-fold>summary::-webkit-details-marker{display:none}
+.ev-fold>summary::before{content:'▸ ';color:var(--text-muted)}
+.ev-fold[open]>summary::before{content:'▾ '}
+.ev-fold[open]>summary{border-bottom:1px solid var(--border);color:var(--text-primary)}
+.ev-fold>summary:hover{color:var(--text-primary)}
+.ev-fold-body{padding:8px 16px 16px}
+.ev-fold-body>section:first-child,.ev-fold-body>.audit-fingerprints:first-child{margin-top:0}
+.ev-fold-body .meta-tags{margin-top:8px}
+/* 逐用例 section 标题 */
+.ev-samples{margin-top:28px}
+.ev-samples-h{font-size:1.4rem;margin:0 0 4px;color:var(--text-primary)}
+.ev-samples-sub{font-size:13.5px;color:var(--text-secondary);margin:0 0 8px;line-height:1.6}
+/* Context strip — 实验配置 + Agent 概览并排,辅助元数据 */
 .ctx-strip { display:grid;gap:14px;grid-template-columns:1fr 1fr;margin:14px 0 24px }
 @media (max-width:760px) { .ctx-strip { grid-template-columns:1fr } }
 .ctx-block { background:var(--bg-surface);border-radius:var(--radius);box-shadow:var(--shadow-sm);padding:12px 16px }
@@ -543,26 +539,30 @@ export function renderRunDetail(report, lang = DEFAULT_LANG) {
 .ctx-row { padding:8px 12px;background:var(--bg-soft);border-radius:4px;display:flex;align-items:center;gap:10px;font-size:13px;line-height:1.5;flex-wrap:wrap }
 .ctx-row-name { font-weight:600;color:var(--text-primary);font-family:"SF Mono",Menlo,monospace;font-size:12.5px;letter-spacing:-0.01em }
 .ctx-row-bits { color:var(--text-secondary);flex:1;min-width:0;font-size:12.5px }
-.ctx-sep { color:var(--text-muted);margin:0 2px;opacity:0.6 }
-/* 报告身份带 — 极简一行,run-id pill + 时间戳 + verdict pill。完整 H1 + meta 移进各 tab 自己的 panel。 */
-.report-id-line { display:flex;align-items:center;gap:12px;flex-wrap:wrap;margin:8px 0 18px }
-.report-id-pill { font-size:14px;font-family:"SF Mono",Menlo,monospace;color:var(--text-primary);background:var(--bg-surface);padding:5px 12px;border-radius:14px;border:1px solid var(--border);letter-spacing:-0.01em;font-weight:500 }
-.report-id-stamp { font-size:12px;color:var(--text-muted);font-variant-numeric:tabular-nums }
-/* 顶部 tab 切换器 — 显式 reset border-radius / appearance / outline,
-   防止全局 button 规则给底部加奇怪圆角和深色背景。 */
-.omk-view-tabs { display:flex;gap:0;margin:16px 0 24px;border-bottom:1px solid var(--border) }
-.omk-view-tab { padding:12px 20px;background:transparent;border:none;border-radius:0;cursor:pointer;color:var(--text-secondary);border-bottom:2px solid transparent;font-size:14px;font-weight:500;outline:none;appearance:none;-webkit-appearance:none;margin-bottom:-1px;font-family:inherit }
-.omk-view-tab--active { color:var(--text-primary);border-bottom-color:var(--accent);font-weight:600 }
-.omk-view-tab:hover:not(.omk-view-tab--active) { color:var(--text-primary) }
-.omk-view-tab:focus-visible { outline:2px solid var(--accent);outline-offset:-4px;border-radius:4px }
-/* tab 面板内的 H1 — 跟列表 H1 同级别但加 emoji 区分视角 */
-.omk-view-panel h1 { font-size:1.5rem;margin:0 0 6px }
-.omk-view-panel .subtitle { font-size:14px;color:var(--text-secondary);margin:0 0 20px;line-height:1.6 }
-    </style>
-    <script>${TEST_VIEW_JS}</script>
-
-    </main>
-  `, lang);
+.ctx-sep { color:var(--text-muted);margin:0 2px;opacity:0.6 }`;
+    return reportShell({
+        dim: 'eval',
+        icon: '🧪',
+        kindTitle: lang === 'zh' ? '用例评测报告' : 'Eval Report',
+        // 头部用规范 skill 名(skill-index entry),与 doctor/observe 一致;无 context 时回退到 report 推导名。
+        skillName: skillContext?.skillName ?? heroSkillName,
+        metaItems: [`${icon('clock', { size: 13 })} ${e(heroTs)}`, `${icon('chip', { size: 13 })} ${e(m.model)}`, e(report.id)],
+        // Hero 展示「综合健康」总分(跨 doctor/eval/observe,与首页同口径),切 tab 不变;
+        // 本次 eval 的综合分(4.09/5)在下方结论卡。无 skillContext(batch/孤儿报告)时回退显 eval 综合分。
+        score: skillContext ? skillContext.overall.score : heroScore100,
+        scoreText: skillContext
+            ? (skillContext.overall.score != null ? String(skillContext.overall.score) : undefined)
+            : (heroComposite != null ? heroComposite.toFixed(2) : undefined),
+        ringColor: skillContext
+            ? healthColor(skillContext.overall.score)
+            : (heroComposite != null ? (heroComposite >= 4 ? '#1f9d63' : heroComposite >= 3 ? '#d97706' : '#dc2626') : undefined),
+        scoreLabel: skillContext ? (lang === 'zh' ? '综合健康' : 'Health') : (lang === 'zh' ? '综合分 / 5' : 'Composite / 5'),
+        backHref: `/${langQ}`,
+        backLabel: lang === 'zh' ? '返回列表' : 'Back to list',
+        body,
+        skillContext,
+        extraCss: evalCss,
+    }, lang);
 }
 export function renderBatchEvaluationDetail(report, lang = DEFAULT_LANG) {
     const langQ = lang === DEFAULT_LANG ? '' : `?lang=${lang}`;
@@ -649,12 +649,12 @@ export function renderBatchEvaluationDetail(report, lang = DEFAULT_LANG) {
     </main>
   `, lang);
 }
-export function renderReportDocumentDetail(report, lang = DEFAULT_LANG) {
+export function renderReportDocumentDetail(report, lang = DEFAULT_LANG, skillContext) {
     if (!report)
         return renderRunDetail(null, lang);
-    return report.reportKind === 'batch-evaluation'
+    return report.kind === 'batch-evaluation'
         ? renderBatchEvaluationDetail(report, lang)
-        : renderRunDetail(report, lang);
+        : renderRunDetail(report, lang, skillContext);
 }
 export function renderTrendsPage(variantName, runs, lang = DEFAULT_LANG) {
     const body = renderTrendsBody(variantName, runs, lang);

package/dist/renderer/icons.d.ts

@@ -0,0 +1,29 @@
+/**
+ * 共享单色线性 SVG 图标集（替代各处 emoji）。
+ *
+ * 设计：统一 24×24 viewBox、`stroke="currentColor"`、单一描边权重、圆角端点。
+ * 颜色由调用处的 CSS `color` 控制（图标自身不带色），尺寸/粗细可覆盖。
+ * 之所以自带一套而不用图标库：报告是离线自包含 HTML，内联 SVG 无需外链/字体。
+ */
+/** name → viewBox 内的 path/shape 串。 */
+declare const PATHS: Record<string, string>;
+export type IconName = keyof typeof PATHS;
+export interface IconOpts {
+    /** 像素尺寸（宽高相同），默认 16。 */
+    size?: number;
+    /** 额外 class。 */
+    cls?: string;
+    /** 描边权重，默认 1.7。 */
+    sw?: number;
+    /** 内联 style。 */
+    style?: string;
+}
+/** 返回一个自包含 `<svg>` 串。未知 name 返回空串（不抛错，调用处可安全拼接）。 */
+export declare function icon(name: string, opts?: IconOpts): string;
+/** 维度 → 报告类型图标名。 */
+export declare function dimIconName(dim: 'doctor' | 'eval' | 'observe'): IconName;
+/** 品牌 logo 的完整 `<svg>`(给定像素尺寸)。 */
+export declare function brandLogo(size?: number): string;
+/** 品牌 logo 的原始 `<svg>`(无尺寸,供 favicon data URI 编码用)。 */
+export declare const BRAND_LOGO_RAW = "<svg xmlns=\"http://www.w3.org/2000/svg\" viewBox=\"0 0 120 120\" fill=\"none\"><defs><linearGradient id=\"omkAccent\" x1=\"0%\" y1=\"0%\" x2=\"100%\" y2=\"0%\"><stop offset=\"0%\" stop-color=\"#60a5fa\"/><stop offset=\"100%\" stop-color=\"#a78bfa\"/></linearGradient></defs><circle cx=\"60\" cy=\"60\" r=\"56\" fill=\"#0f172a\" stroke=\"#1e293b\" stroke-width=\"1.5\"/><path d=\"M24 60 A12 12 0 1 1 36 72\" stroke=\"url(#omkAccent)\" stroke-width=\"3.5\" stroke-linecap=\"round\" fill=\"none\"/><path d=\"M56 72 V57 Q56 48 61.5 48 Q67 48 67 57 V72 M67 57 Q67 48 72.5 48 Q78 48 78 57 V72\" stroke=\"#e2e8f0\" stroke-width=\"3.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" fill=\"none\"/><path d=\"M86 48 V72 M86 60 L96 48 M86 60 L96 72\" stroke=\"#e2e8f0\" stroke-width=\"3.5\" stroke-linecap=\"round\" stroke-linejoin=\"round\" fill=\"none\"/></svg>";
+export {};