dsh-crwu-workbench 0.0.10

package/skills/crwu/crwu-dev-audit-optimize/scripts/examples/gap-analysis.sample.json

@@ -0,0 +1,116 @@
+{
+  "schemaVersion": "1.0.0",
+  "reportId": "GAP-BG8169-20260915",
+  "generatedAt": "2026-09-15T16:00:00+08:00",
+  "project": {
+    "projectId": "BG8169",
+    "auditResult": "审核意见.BG8169.json",
+    "reviewFiles": ["人工复核意见.xlsx"],
+    "finalMaterials": ["评估报告-最终版.docx", "市场法测算.xlsx"]
+  },
+  "baselines": {
+    "auditSnapshot": {
+      "manifest": "knowledge/.crwu-manifest.jsonl",
+      "verifiedSameProject": true,
+      "exportedAt": "2026-09-15T09:00:00+08:00"
+    },
+    "currentKbSnapshot": {
+      "readOnly": true,
+      "checkedAt": "2026-09-15T15:00:00+08:00"
+    }
+  },
+  "comparison": {
+    "overlap": 6,
+    "aiOnly": 3,
+    "divergent": 1
+  },
+  "summary": {
+    "reviewerOnly": 3,
+    "actionableMisses": 2,
+    "resolved": 1,
+    "uncheckable": 0,
+    "questionable": 0,
+    "repairPlans": 2
+  },
+  "reviewerOnlyItems": [
+    {
+      "gapId": "L-01",
+      "title": "市场比较案例时间修正依据不足",
+      "status": "L-open",
+      "reviewerEvidence": {"file": "人工复核意见.xlsx", "location": "问题清单!B12", "excerpt": "时间修正依据不充分"},
+      "finalDocumentVerification": {"file": "市场法测算.xlsx", "location": "案例表!H8", "result": "仍存在"},
+      "executionTrace": [
+        {"stage": "input", "status": "passed", "evidence": "材料清单#2"},
+        {"stage": "profile", "status": "passed", "evidence": "route_profile.methods=市场法"},
+        {"stage": "routing", "status": "passed", "evidence": "已加载房地产与交易处置"},
+        {"stage": "assembly", "status": "passed", "evidence": "manifest 已下载清单-M-市场法"},
+        {"stage": "knowledge", "status": "failed", "evidence": "CHK-MKT-012 内容不足"}
+      ],
+      "rootCauses": [{"code": "K2", "role": "primary", "evidence": "历史与当前正文均缺判断标准"}],
+      "linkedFixIds": ["FIX-KB-01"]
+    },
+    {
+      "gapId": "L-02",
+      "title": "土地用途限制条件未核验",
+      "status": "L-open",
+      "reviewerEvidence": {"file": "人工复核意见.xlsx", "location": "问题清单!B18", "excerpt": "未核验规划用途限制"},
+      "finalDocumentVerification": {"file": "评估报告-最终版.docx", "location": "第45页", "result": "仍存在"},
+      "executionTrace": [
+        {"stage": "knowledge", "status": "passed", "evidence": "CHK-RE-021 已装载"},
+        {"stage": "skill", "status": "failed", "evidence": "未定义四端交叉核验"},
+        {"stage": "output", "status": "not_applicable", "evidence": "上游未形成 finding"}
+      ],
+      "rootCauses": [{"code": "S2", "role": "primary", "evidence": "知识已有但 Skill 未执行"}],
+      "linkedFixIds": ["FIX-SKILL-01"]
+    },
+    {
+      "gapId": "L-03",
+      "title": "报告未披露土地剩余年限",
+      "status": "L-resolved",
+      "reviewerEvidence": {"file": "人工复核意见.xlsx", "location": "问题清单!B21", "excerpt": "补充剩余年限"},
+      "finalDocumentVerification": {"file": "评估报告-最终版.docx", "location": "第31页", "result": "已整改"},
+      "executionTrace": [],
+      "rootCauses": [{"code": "H", "role": "primary", "evidence": "最终版已落实"}],
+      "linkedFixIds": []
+    }
+  ],
+  "repairPlans": [
+    {
+      "fixId": "FIX-KB-01",
+      "title": "补充时间修正依据充分性检查点",
+      "type": "kb_manual",
+      "executionMode": "manual_only",
+      "approvalState": "pending",
+      "sourceGapIds": ["L-01"],
+      "target": {"kbPath": "06-规则库/评估方法/市场法/清单-M-市场法", "anchor": "CHK-MKT-012"},
+      "changeSpec": {
+        "currentGap": "缺少时间修正依据充分性的证据、步骤、标准和例外",
+        "method": "扩写现有 CHK-MKT-012",
+        "contentOutline": ["适用条件", "必取证据", "检查步骤", "判断标准", "异常输出", "例外"]
+      },
+      "registrations": ["条目计数变化时更新批次 README"],
+      "completionEvidence": "用户确认 FIX-KB-01 已完成并允许只读复核",
+      "dependencies": [],
+      "validation": ["锚点唯一", "原漏检案例回归"]
+    },
+    {
+      "fixId": "FIX-SKILL-01",
+      "title": "补充土地用途限制四端交叉核验",
+      "type": "skill_ai",
+      "executionMode": "ai_after_approval",
+      "approvalState": "approved",
+      "sourceGapIds": ["L-02"],
+      "target": {"skillFile": "crwu-audit-asset-realestate/references/02-review-focus.md", "anchor": "CHK-RE-021"},
+      "changeSpec": {
+        "currentGap": "未定义权证、规划条件、报告披露、测算假设四端交叉核验",
+        "method": "补充执行步骤和证据要求",
+        "contentOutline": ["四端取证", "冲突判断", "输出证据"]
+      },
+      "registrations": ["源仓变更纪要"],
+      "completionEvidence": "相关测试和原漏检回归通过",
+      "dependencies": [],
+      "validation": ["quick_validate", "kb_tool validate", "原漏检案例回归"]
+    }
+  ],
+  "renderPolicy": {"offline": true, "printA4": true}
+}

package/skills/crwu/crwu-dev-audit-optimize/scripts/gap_analysis.schema.json

@@ -0,0 +1,33 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://crwu.local/schema/gap-analysis/1.0.0",
+  "title": "GapAnalysis",
+  "type": "object",
+  "required": [
+    "schemaVersion",
+    "reportId",
+    "project",
+    "baselines",
+    "summary",
+    "reviewerOnlyItems",
+    "repairPlans",
+    "renderPolicy"
+  ],
+  "properties": {
+    "schemaVersion": {"type": "string", "pattern": "^1\\."},
+    "reportId": {"type": "string", "pattern": "^GAP-[A-Za-z0-9_-]+-[0-9]{8}$"},
+    "project": {"type": "object"},
+    "baselines": {"type": "object"},
+    "summary": {"type": "object"},
+    "reviewerOnlyItems": {"type": "array", "items": {"type": "object"}},
+    "repairPlans": {"type": "array", "items": {"type": "object"}},
+    "renderPolicy": {
+      "type": "object",
+      "required": ["offline", "printA4"],
+      "properties": {
+        "offline": {"const": true},
+        "printA4": {"const": true}
+      }
+    }
+  }
+}

package/skills/crwu/crwu-dev-audit-optimize/scripts/gap_analysis_delivery.py

@@ -0,0 +1,475 @@
+#!/usr/bin/env python3
+"""GapAnalysis JSON validation and deterministic HTML delivery."""
+
+from __future__ import annotations
+
+import argparse
+import html
+import json
+import re
+import sys
+from pathlib import Path
+
+
+ACTIONABLE = {"L-open", "L-unclosed"}
+NON_ACTIONABLE = {"L-resolved", "L-uncheckable", "L-questionable"}
+CAUSES = {"K1", "K2", "S1", "S2", "R", "I", "E", "O", "H"}
+PLAN_TYPES = {"kb_manual", "skill_ai", "maintainer_handoff", "no_action"}
+TRACE_STAGES = {"input", "profile", "routing", "assembly", "knowledge", "skill", "execution", "output"}
+TRACE_STATUSES = {"passed", "failed", "unknown", "not_applicable"}
+ABSOLUTE_PATH = re.compile(r"(?:^|\s)(?:/Users/|/home/|[A-Za-z]:\\\\)")
+SKILL_ROOT = Path(__file__).resolve().parent.parent
+TEMPLATE_PATH = SKILL_ROOT / "template" / "gap-analysis-report.html"
+ECHARTS_PATH = SKILL_ROOT / "template" / "echarts.min.js"
+
+
+def load_result(path: Path) -> dict:
+    with path.open("r", encoding="utf-8") as handle:
+        value = json.load(handle)
+    if not isinstance(value, dict):
+        raise ValueError("GapAnalysis 顶层必须是对象")
+    return value
+
+
+def _unique_index(items: list, key: str, label: str, errors: list[str]) -> dict:
+    result = {}
+    for index, item in enumerate(items):
+        if not isinstance(item, dict):
+            errors.append(f"{label}[{index}] 必须是对象")
+            continue
+        value = item.get(key)
+        if not isinstance(value, str) or not value.strip():
+            errors.append(f"{label}[{index}].{key} 不得为空")
+            continue
+        if value in result:
+            errors.append(f"{label}.{key} 重复：{value}")
+            continue
+        result[value] = item
+    return result
+
+
+def _iter_scalars(value):
+    if isinstance(value, dict):
+        for item in value.values():
+            yield from _iter_scalars(item)
+    elif isinstance(value, list):
+        for item in value:
+            yield from _iter_scalars(item)
+    elif isinstance(value, str):
+        yield value
+
+
+def _scan_forbidden_content(result: dict) -> list[str]:
+    errors = []
+    for value in _iter_scalars(result):
+        if ABSOLUTE_PATH.search(value):
+            errors.append("报告不得包含个人绝对路径")
+            break
+    return errors
+
+
+def _validate_summary(result: dict, gaps: list, plans: list) -> list[str]:
+    summary = result.get("summary") or {}
+    counts = {
+        "reviewerOnly": len(gaps),
+        "actionableMisses": sum(item.get("status") in ACTIONABLE for item in gaps if isinstance(item, dict)),
+        "resolved": sum(item.get("status") == "L-resolved" for item in gaps if isinstance(item, dict)),
+        "uncheckable": sum(item.get("status") == "L-uncheckable" for item in gaps if isinstance(item, dict)),
+        "questionable": sum(item.get("status") == "L-questionable" for item in gaps if isinstance(item, dict)),
+        "repairPlans": len(plans),
+    }
+    return [f"summary.{key} 必须可由明细重算" for key, expected in counts.items() if summary.get(key) != expected]
+
+
+def validate(result: dict) -> list[str]:
+    errors: list[str] = []
+    if not isinstance(result, dict):
+        return ["GapAnalysis 顶层必须是对象"]
+    required = (
+        "schemaVersion", "reportId", "project", "baselines", "summary",
+        "reviewerOnlyItems", "repairPlans", "renderPolicy",
+    )
+    for key in required:
+        if key not in result:
+            errors.append(f"缺少顶层字段 {key}")
+
+    gaps = result.get("reviewerOnlyItems") or []
+    plans = result.get("repairPlans") or []
+    if not isinstance(gaps, list):
+        errors.append("reviewerOnlyItems 必须是数组")
+        gaps = []
+    if not isinstance(plans, list):
+        errors.append("repairPlans 必须是数组")
+        plans = []
+    gap_by_id = _unique_index(gaps, "gapId", "reviewerOnlyItems", errors)
+    plan_by_id = _unique_index(plans, "fixId", "repairPlans", errors)
+
+    for gap_id, gap in gap_by_id.items():
+        status = gap.get("status")
+        linked = gap.get("linkedFixIds") or []
+        if status in ACTIONABLE and not linked:
+            errors.append(f"{gap_id}.linkedFixIds：确认漏检必须关联修复计划")
+        for cause in gap.get("rootCauses") or []:
+            if not isinstance(cause, dict) or cause.get("code") not in CAUSES:
+                errors.append(f"{gap_id}.rootCauses 包含非法根因")
+        for trace in gap.get("executionTrace") or []:
+            if not isinstance(trace, dict) or trace.get("stage") not in TRACE_STAGES:
+                errors.append(f"{gap_id}.executionTrace 包含非法阶段")
+            elif trace.get("status") not in TRACE_STATUSES:
+                errors.append(f"{gap_id}.executionTrace 包含非法状态")
+        for fix_id in linked:
+            if fix_id not in plan_by_id:
+                errors.append(f"{gap_id}.linkedFixIds 引用未知计划 {fix_id}")
+            elif gap_id not in (plan_by_id[fix_id].get("sourceGapIds") or []):
+                errors.append(f"{gap_id} 与 {fix_id} 双向关系不一致")
+
+    for fix_id, plan in plan_by_id.items():
+        sources = plan.get("sourceGapIds") or []
+        if not sources:
+            errors.append(f"{fix_id}.sourceGapIds 不得为空")
+        if plan.get("type") not in PLAN_TYPES:
+            errors.append(f"{fix_id}.type 非法")
+        if plan.get("type") == "kb_manual" and plan.get("executionMode") != "manual_only":
+            errors.append(f"{fix_id} 知识库计划必须为 manual_only")
+        for gap_id in sources:
+            gap = gap_by_id.get(gap_id)
+            if not gap:
+                errors.append(f"{fix_id}.sourceGapIds 引用未知漏检 {gap_id}")
+            elif gap.get("status") in NON_ACTIONABLE:
+                errors.append(f"{fix_id} 不得包含 {gap.get('status')} 项 {gap_id}")
+            elif fix_id not in (gap.get("linkedFixIds") or []):
+                errors.append(f"{fix_id} 与 {gap_id} 双向关系不一致")
+
+    errors.extend(_validate_summary(result, gaps, plans))
+    errors.extend(_scan_forbidden_content(result))
+    policy = result.get("renderPolicy") or {}
+    if policy.get("offline") is not True or policy.get("printA4") is not True:
+        errors.append("renderPolicy 必须要求离线单文件和 A4 打印")
+    return errors
+
+
+def _h(value) -> str:
+    return html.escape("" if value is None else str(value), quote=True)
+
+
+def build_manual_kb_instruction(plan: dict) -> str:
+    if plan.get("type") != "kb_manual" or plan.get("executionMode") != "manual_only":
+        raise ValueError("仅 kb_manual/manual_only 可生成人工修复单")
+    target = plan.get("target") or {}
+    change = plan.get("changeSpec") or {}
+    return "\n".join(
+        [
+            f"知识库人工修复单 {plan.get('fixId', '')}（仅人工）",
+            f"对应人工独有事项：{'、'.join(plan.get('sourceGapIds') or [])}",
+            f"目标文档：{target.get('kbPath', '')}",
+            f"定位锚点：{target.get('anchor', '')}",
+            f"当前缺口：{change.get('currentGap', '')}",
+            f"修改方式：{change.get('method', '')}",
+            f"建议内容结构：{'；'.join(change.get('contentOutline') or [])}",
+            f"联动登记：{'；'.join(plan.get('registrations') or [])}",
+            f"完成凭证：{plan.get('completionEvidence', '')}",
+        ]
+    )
+
+
+def build_skill_prompt(result: dict, approved_fix_ids: list[str]) -> str:
+    plan_by_id = {item.get("fixId"): item for item in result.get("repairPlans") or []}
+    if not approved_fix_ids:
+        raise ValueError("未批准任何 Skill 修复计划")
+    selected = []
+    for fix_id in approved_fix_ids:
+        plan = plan_by_id.get(fix_id)
+        if not plan:
+            raise ValueError(f"未知修复计划 {fix_id}")
+        if plan.get("type") != "skill_ai" or plan.get("approvalState") != "approved":
+            raise ValueError(f"计划 {fix_id} 未批准或不是 Skill AI 计划")
+        selected.append(plan)
+
+    lines = [
+        "我确认执行《AI—人工审核差距分析报告 {0}》中的 Skill 修复计划：{1}。".format(
+            result.get("reportId", ""), "、".join(approved_fix_ids)
+        ),
+        "",
+        "请使用 crwu-dev-audit-optimize 执行上述已批准计划。先核对报告证据、源仓当前状态和依赖；涉及知识库依赖时，只允许通过 crwu-dws 只读重新下载核验，不得对知识库执行任何写操作。",
+        "",
+        "执行要求：",
+        "1. 仅修改以下批准计划列明的源仓 Skill 文件，不修改运行时目录，不扩大范围。",
+    ]
+    for number, plan in enumerate(selected, 2):
+        target = plan.get("target") or {}
+        change = plan.get("changeSpec") or {}
+        lines.append(
+            f"{number}. {plan['fixId']}（对应 {'、'.join(plan.get('sourceGapIds') or [])}）："
+            f"在 {target.get('skillFile', '')} 的 {target.get('anchor', '')} 处{change.get('method', '')}；"
+            f"目标内容为 {'、'.join(change.get('contentOutline') or [])}。"
+        )
+    next_number = len(selected) + 2
+    lines.extend(
+        [
+            f"{next_number}. 只写 RULE/CHK 编号和库内层级路径寻址键，不复制知识库正文。",
+            f"{next_number + 1}. 若最新状态冲突、目标路径变化、依赖未满足或需要修改清单外文件，立即停止并说明。",
+            f"{next_number + 2}. 完成后运行各计划 validation 中的校验，并按 sourceGapIds 逐条回归原人工漏检，汇报文件级改动、测试结果和剩余风险。",
+        ]
+    )
+    return "\n".join(lines)
+
+
+def actionable_cause_counts(result: dict) -> dict[str, int]:
+    counts: dict[str, int] = {}
+    for gap in result.get("reviewerOnlyItems") or []:
+        if gap.get("status") not in ACTIONABLE:
+            continue
+        for cause in gap.get("rootCauses") or []:
+            code = cause.get("code")
+            if code:
+                counts[code] = counts.get(code, 0) + 1
+    return counts
+
+
+def _badge(text: str) -> str:
+    return f'<span class="badge">{_h(text)}</span>'
+
+
+def _render_summary(result: dict) -> str:
+    summary = result.get("summary") or {}
+    comparison = result.get("comparison") or {}
+    stats = [
+        ("双方共同发现", comparison.get("overlap", 0)),
+        ("AI 新增", comparison.get("aiOnly", 0)),
+        ("人工独有", summary.get("reviewerOnly", 0)),
+        ("确认 AI 漏检", summary.get("actionableMisses", 0)),
+        ("已整改", summary.get("resolved", 0)),
+        ("不可核验", summary.get("uncheckable", 0)),
+        ("人工意见存疑", summary.get("questionable", 0)),
+        ("修复计划", summary.get("repairPlans", 0)),
+    ]
+    cards = "".join(f'<div class="stat"><span>{_h(label)}</span><strong>{_h(value)}</strong></div>' for label, value in stats)
+    return f'<section id="gap-summary"><h2>差距摘要</h2><div class="stats">{cards}</div></section>'
+
+
+def _render_gap_rows(result: dict) -> str:
+    rows = []
+    plans = {item.get("fixId"): item for item in result.get("repairPlans") or []}
+    for gap in result.get("reviewerOnlyItems") or []:
+        causes = "、".join(item.get("code", "") for item in gap.get("rootCauses") or [])
+        linked = gap.get("linkedFixIds") or []
+        targets = []
+        for fix_id in linked:
+            target = (plans.get(fix_id) or {}).get("target") or {}
+            targets.append(target.get("kbPath") or target.get("skillFile") or "")
+        verification = gap.get("finalDocumentVerification") or {}
+        rows.append(
+            '<tr data-gap-id="{gap_id}"><td><strong>{gap_id}</strong><br>{title}</td>'
+            '<td>{status}<br><span class="muted">{result}</span></td><td>{causes}</td>'
+            '<td>{fixes}</td><td>{targets}</td></tr>'.format(
+                gap_id=_h(gap.get("gapId")),
+                title=_h(gap.get("title")),
+                status=_badge(gap.get("status", "")),
+                result=_h(verification.get("result", "")),
+                causes=_h(causes),
+                fixes=" ".join(_badge(item) for item in linked) or "—",
+                targets="<br>".join(_h(item) for item in targets) or "—",
+            )
+        )
+    return "".join(rows)
+
+
+def _render_traces(result: dict) -> str:
+    blocks = []
+    for gap in result.get("reviewerOnlyItems") or []:
+        if gap.get("status") not in ACTIONABLE:
+            continue
+        trace = []
+        for index, item in enumerate(gap.get("executionTrace") or []):
+            if index:
+                trace.append('<span class="trace-line" aria-hidden="true"></span>')
+            trace.append(
+                '<span class="trace-step {status}" title="{evidence}">{stage} · {status}</span>'.format(
+                    status=_h(item.get("status")), stage=_h(item.get("stage")), evidence=_h(item.get("evidence"))
+                )
+            )
+        blocks.append(
+            '<div data-gap-id="{gap_id}"><h3>{gap_id} · {title}</h3><div class="trace">{trace}</div></div>'.format(
+                gap_id=_h(gap.get("gapId")), title=_h(gap.get("title")), trace="".join(trace) or "无执行链记录"
+            )
+        )
+    return "".join(blocks) or '<p class="muted">没有需要归因的真实漏检。</p>'
+
+
+def _render_manual_plans(result: dict) -> str:
+    cards = []
+    for plan in result.get("repairPlans") or []:
+        if plan.get("type") != "kb_manual":
+            continue
+        fix_id = plan.get("fixId", "")
+        instruction = build_manual_kb_instruction(plan)
+        cards.append(
+            '<article class="plan"><h3>{fix_id} · {title}</h3><p class="manual-only">manual_only · 仅人工修改</p>'
+            '<pre id="manual-{fix_id}">{instruction}</pre><div class="copy-row">'
+            '<button type="button" data-copy-target="manual-{fix_id}">复制人工修复说明</button><span class="muted copy-status"></span>'
+            '</div></article>'.format(fix_id=_h(fix_id), title=_h(plan.get("title")), instruction=_h(instruction))
+        )
+    return "".join(cards) or '<p class="muted">本次没有知识库人工修复项。</p>'
+
+
+def _render_skill_plans(result: dict) -> str:
+    approved = [item.get("fixId") for item in result.get("repairPlans") or [] if item.get("type") == "skill_ai" and item.get("approvalState") == "approved"]
+    cards = []
+    for plan in result.get("repairPlans") or []:
+        if plan.get("type") != "skill_ai":
+            continue
+        target = plan.get("target") or {}
+        cards.append(
+            '<article class="plan skill"><h3>{fix_id} · {title}</h3><dl>'
+            '<dt>对应漏检</dt><dd>{gaps}</dd><dt>源仓文件</dt><dd><code>{file}</code></dd>'
+            '<dt>修改方式</dt><dd>{method}</dd><dt>批准状态</dt><dd>{approval}</dd></dl></article>'.format(
+                fix_id=_h(plan.get("fixId")), title=_h(plan.get("title")), gaps=_h("、".join(plan.get("sourceGapIds") or [])),
+                file=_h(target.get("skillFile")), method=_h((plan.get("changeSpec") or {}).get("method")), approval=_h(plan.get("approvalState")),
+            )
+        )
+    prompt = build_skill_prompt(result, approved) if approved else "没有已批准的 Skill 修复计划。"
+    cards.append(
+        '<article class="plan skill"><h3>可复制给 AI 的执行 Prompt</h3>'
+        '<textarea id="skill-execution-prompt" readonly>{prompt}</textarea><div class="copy-row">'
+        '<button type="button" class="primary" data-copy-target="skill-execution-prompt">复制 Skill 执行 Prompt</button>'
+        '<span class="muted copy-status"></span></div></article>'.format(prompt=_h(prompt))
+    )
+    return "".join(cards)
+
+
+def _render_validation(result: dict) -> str:
+    rows = []
+    for plan in result.get("repairPlans") or []:
+        rows.append(
+            "<tr><td>{}</td><td>{}</td><td>{}</td><td>{}</td></tr>".format(
+                _h(plan.get("fixId")), _h("、".join(plan.get("sourceGapIds") or [])),
+                _h("、".join(plan.get("dependencies") or []) or "无"), _h("；".join(plan.get("validation") or [])),
+            )
+        )
+    return "".join(rows)
+
+
+def _render_sections(result: dict) -> str:
+    project = result.get("project") or {}
+    return "".join(
+        [
+            '<header><h1>AI—人工审核差距分析</h1><p class="subtitle">项目 {project} · 报告 {report}</p>'
+            '<p>历史基线用于还原当时漏检原因；当前知识库仅通过 crwu-dws 只读核验缺口是否仍存在。</p></header>'.format(
+                project=_h(project.get("projectId")), report=_h(result.get("reportId"))
+            ),
+            _render_summary(result),
+            '<section id="gap-charts"><h2>差距构成与根因分布</h2><div class="charts">'
+            '<div id="composition-chart" class="chart" aria-label="差距构成图"></div>'
+            '<div id="cause-chart" class="chart" aria-label="根因分布图"></div></div></section>',
+            '<section id="reviewer-only-table"><h2>人工独有事项逐条归因</h2><div class="table-wrap"><table>'
+            '<thead><tr><th>人工独有事项</th><th>在件核验</th><th>根因</th><th>对应修复计划</th><th>修复目标</th></tr></thead>'
+            f'<tbody>{_render_gap_rows(result)}</tbody></table></div></section>',
+            f'<section id="execution-trace"><h2>真实漏检执行链</h2>{_render_traces(result)}</section>',
+            '<section id="gap-fix-map"><h2>人工独有事项 → 根因 → 修复计划</h2>'
+            '<div id="relation-chart" aria-label="漏检与修复计划关系图"></div></section>',
+            f'<section id="manual-kb-plans"><h2>知识库人工修复单</h2><p class="manual-only">知识库只能由用户人工修改，AI 不得写入。</p><div class="plan-grid">{_render_manual_plans(result)}</div></section>',
+            f'<section id="skill-repair-prompts"><h2>Skill 修复计划与执行 Prompt</h2><div class="plan-grid">{_render_skill_plans(result)}</div></section>',
+            '<section id="validation-plan"><h2>依赖与验证</h2><div class="table-wrap"><table><thead><tr>'
+            '<th>计划</th><th>对应漏检</th><th>依赖</th><th>验证</th></tr></thead>'
+            f'<tbody>{_render_validation(result)}</tbody></table></div></section>',
+            '<footer>由 crwu-dev-audit-optimize 差距分析模式生成 · 表格是证据主载体，图表仅用于汇总与导航</footer>',
+        ]
+    )
+
+
+def _report_javascript() -> str:
+    return r"""
+(function(){
+  const data = JSON.parse(document.getElementById('gap-analysis-data').textContent);
+  const colors = ['#356da8','#9a5b08','#7652a8','#a92d28','#39724f','#667085'];
+  const gaps = data.reviewerOnlyItems || [];
+  const plans = data.repairPlans || [];
+  const causeCounts = data.chartData.actionableCauseCounts;
+  const composition = echarts.init(document.getElementById('composition-chart'));
+  composition.setOption({animation:false,tooltip:{trigger:'axis'},grid:{left:100,right:25,top:20,bottom:30},
+    xAxis:{type:'value'},yAxis:{type:'category',data:['双方共同','AI 新增','结论分歧','人工独有','确认漏检']},
+    series:[{type:'bar',data:[data.comparison.overlap,data.comparison.aiOnly,data.comparison.divergent,data.summary.reviewerOnly,data.summary.actionableMisses],label:{show:true,position:'right'},itemStyle:{color:'#356da8'}}]});
+  const causes = echarts.init(document.getElementById('cause-chart'));
+  causes.setOption({animation:false,tooltip:{trigger:'item'},legend:{bottom:0},series:[{type:'pie',radius:['42%','68%'],data:Object.entries(causeCounts).map(([name,value],i)=>({name,value,itemStyle:{color:colors[i%colors.length]}}))}]});
+  const nodes = [], links = [], seen = new Set();
+  const addNode = (name, depth) => { const key=depth+':'+name; if(!seen.has(key)){seen.add(key);nodes.push({name:key,label:{formatter:name},depth});} return key; };
+  gaps.filter(g=>['L-open','L-unclosed'].includes(g.status)).forEach(g=>{
+    const left=addNode(g.gapId,0);
+    (g.rootCauses||[]).forEach(c=>{const middle=addNode(c.code,1);links.push({source:left,target:middle,value:1});
+      (g.linkedFixIds||[]).forEach(id=>{const right=addNode(id,2);if(!links.some(x=>x.source===middle&&x.target===right))links.push({source:middle,target:right,value:1});});
+    });
+  });
+  const relation = echarts.init(document.getElementById('relation-chart'));
+  relation.setOption({animation:false,tooltip:{trigger:'item'},series:[{type:'sankey',left:18,right:90,nodeWidth:15,nodeGap:13,draggable:false,data:nodes,links,lineStyle:{color:'gradient',opacity:.35}}]});
+  function focusGap(gapId){document.querySelectorAll('[data-gap-id]').forEach(el=>{el.style.opacity=!gapId||el.dataset.gapId===gapId?'1':'.28';});}
+  document.querySelectorAll('[data-gap-id]').forEach(el=>el.addEventListener('click',()=>focusGap(el.dataset.gapId)));
+  relation.on('click',p=>{const label=(p.data&&p.data.label&&p.data.label.formatter)||'';if(/^L-/.test(label))focusGap(label);});
+  document.querySelectorAll('[data-copy-target]').forEach(button=>button.addEventListener('click',async()=>{
+    const target=document.getElementById(button.dataset.copyTarget);const text='value' in target?target.value:target.textContent;const status=button.parentElement.querySelector('.copy-status');
+    try{await navigator.clipboard.writeText(text);status.textContent='已复制';}catch(_){if(target.select)target.select();status.textContent='请按 Ctrl/Cmd+C 复制';}
+  }));
+  window.addEventListener('resize',()=>{composition.resize();causes.resize();relation.resize();});
+})();
+"""
+
+
+def render(result: dict) -> str:
+    errors = validate(result)
+    if errors:
+        raise ValueError("GapAnalysis 校验失败：\n" + "\n".join(errors))
+    rendered_result = json.loads(json.dumps(result, ensure_ascii=False))
+    rendered_result["chartData"] = {"actionableCauseCounts": actionable_cause_counts(result)}
+    payload = json.dumps(rendered_result, ensure_ascii=False, sort_keys=True, separators=(",", ":"))
+    payload = payload.replace("<", "\\u003c").replace(">", "\\u003e").replace("&", "\\u0026")
+    template = TEMPLATE_PATH.read_text(encoding="utf-8")
+    values = {
+        "{{report_title}}": _h(f"AI—人工审核差距分析 · {(result.get('project') or {}).get('projectId', '')}"),
+        "{{report_content}}": _render_sections(result),
+        "{{report_payload}}": payload,
+        "{{echarts_bundle}}": ECHARTS_PATH.read_text(encoding="utf-8"),
+        "{{report_javascript}}": _report_javascript(),
+    }
+    for key, value in values.items():
+        template = template.replace(key, value)
+    if re.search(r"\{\{[a-z_]+\}\}", template):
+        raise ValueError("HTML 模板仍有未替换占位符")
+    return template
+
+
+def _cmd_validate(args) -> int:
+    result = load_result(Path(args.input))
+    errors = validate(result)
+    if errors:
+        for error in errors:
+            print(f"ERROR: {error}", file=sys.stderr)
+        return 1
+    print("OK")
+    return 0
+
+
+def _cmd_render(args) -> int:
+    result = load_result(Path(args.input))
+    document = render(result)
+    output = Path(args.output)
+    output.parent.mkdir(parents=True, exist_ok=True)
+    output.write_text(document, encoding="utf-8")
+    print(output)
+    return 0
+
+
+def main(argv=None) -> int:
+    parser = argparse.ArgumentParser(description="CRWU AI—人工审核差距分析校验与单文件 HTML 渲染")
+    subparsers = parser.add_subparsers(dest="command", required=True)
+    validate_parser = subparsers.add_parser("validate", help="校验 GapAnalysis JSON")
+    validate_parser.add_argument("--input", required=True)
+    validate_parser.set_defaults(func=_cmd_validate)
+    render_parser = subparsers.add_parser("render", help="渲染自包含单文件 HTML")
+    render_parser.add_argument("--input", required=True)
+    render_parser.add_argument("--output", required=True)
+    render_parser.set_defaults(func=_cmd_render)
+    args = parser.parse_args(argv)
+    return args.func(args)
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())