maintain | 2026-08-07 D 证据核验通过 + frontmatter 合规修复（type/confidence 补齐 23 处）

2026-08-07 17:32:15 +08:00
parent bc91c6ba07
commit 87f8143db0
25 changed files with 57 additions and 6 deletions
@@ -110,10 +110,10 @@ status: generated
 ## 超长页面
 - `wiki/Glossary.md`：695 行；先判断是否应按职责拆分，不以行数为唯一删除依据
- `wiki/concepts/Meta-Harness.md`：328 行；先判断是否应按职责拆分，不以行数为唯一删除依据
+- `wiki/concepts/Meta-Harness.md`：330 行；先判断是否应按职责拆分，不以行数为唯一删除依据
- `wiki/concepts/Harness-Engineering-Complete-Guide.md`：324 行；先判断是否应按职责拆分，不以行数为唯一删除依据
+- `wiki/concepts/Harness-Engineering-Complete-Guide.md`：326 行；先判断是否应按职责拆分，不以行数为唯一删除依据
- `wiki/queries/What-is-Harness-Engineering-in-Simple-Terms.md`：290 行；先判断是否应按职责拆分，不以行数为唯一删除依据
+- `wiki/queries/What-is-Harness-Engineering-in-Simple-Terms.md`：292 行；先判断是否应按职责拆分，不以行数为唯一删除依据
- `wiki/practices/OpenAI-Codex-Harness-Engineering.md`：261 行；先判断是否应按职责拆分，不以行数为唯一删除依据
+- `wiki/practices/OpenAI-Codex-Harness-Engineering.md`：263 行；先判断是否应按职责拆分，不以行数为唯一删除依据
 - `wiki/INDEX.md`：202 行；先判断是否应按职责拆分，不以行数为唯一删除依据
 ## 缺少页面定位区块
@@ -410,3 +410,9 @@
 - D 第一批：20 个 low_confidence 页面核验——全部为作者观点/历史规则/二手来源/时点数据类，low 标注与页面自声明一致，维持不变；62 个 contested 抽查确认均有版本/时点/来源说明，标注完整，维持不变。
 - A：Harness 概念统一定义收敛到核心页 Harness-Engineering.md（「什么是 Harness？（统一定义）」节声明为全库定义源），各来源案例页定位区块已指向核心页。
 - 验证：wiki-reflect.ps1 复核 missing_role 0；orphans 0；duplicate_terms 0。
 ## [2026-08-07] maintain | D 第二批核验 + frontmatter 合规修复
 - D 第二批：62 个 contested 页面逐页信号核验——全部为 A 类（标注完整：均有版本/年份/时点/核验声明，页面定位明确证据边界），B 类（需补标注）0、C 类（可疑）0，维持不变。累计 20 low_confidence + 62 contested 全部核验完成，无页面需升级/降级。
 - 顺带修复 frontmatter 合规：23 处。利弗莫尔交易纪律 type=practices→practice；22 个 llm-wiki 初始批次页面补 type（14 concept / 8 practice / 1 query）与 confidence: medium；paddleocr-ocr工具链 sources 键改名 raw_sources 并补 last_updated（初插位置破坏 YAML 结构，已修复并全库 PyYAML 校验通过）。
 - 验证：wiki-audit.ps1 normal + strict 双绿（0 ERROR / 0 WARNING）；reflect missing_role=0 保持；97 页 frontmatter 全部 YAML 可解析。
@@ -1,5 +1,7 @@
 ---
 title: "智能体协议"
 type: concept
 confidence: medium
 source: "Externalization in LLM Agents: A Unified Review of Memory, Skills, Protocols and Harness Engineering"
 raw_sources:
  - path: raw/Externalization in LLM Agents A Unified Review of Memory, Skills, Protocols and Harness Engineering.md
@@ -1,5 +1,7 @@
 ---
 title: "上下文预算管理"
 type: concept
 confidence: medium
 source: "Externalization in LLM Agents: A Unified Review; nashsu/llm_wiki README"
 raw_sources:
  - path: raw/Externalization in LLM Agents A Unified Review of Memory, Skills, Protocols and Harness Engineering.md
@@ -1,5 +1,7 @@
 ---
 title: "LLM Agent 中的外部化"
 type: concept
 confidence: medium
 source: "Externalization in LLM Agents: A Unified Review of Memory, Skills, Protocols and Harness Engineering"
 raw_sources:
  - path: raw/Externalization in LLM Agents A Unified Review of Memory, Skills, Protocols and Harness Engineering.md
@@ -1,5 +1,7 @@
 ---
 title: "从权重到上下文到 Harness"
 type: concept
 confidence: medium
 source: "Externalization in LLM Agents: A Unified Review of Memory, Skills, Protocols and Harness Engineering"
 raw_sources:
  - path: raw/Externalization in LLM Agents A Unified Review of Memory, Skills, Protocols and Harness Engineering.md
@@ -1,5 +1,7 @@
 ---
 title: "Harness 工程完整指南"
 type: concept
 confidence: medium
 source: "https://www.nxcode.io/resources/news/harness-engineering-complete-guide-ai-agent-codex-2026"
 author: "NxCode Team"
 published: 2026-03-01
@@ -1,5 +1,7 @@
 ---
 title: "Harness 工程：最初的思考"
 type: concept
 confidence: medium
 source: "https://martinfowler.com/articles/exploring-gen-ai/harness-engineering-memo.html"
 author: "Birgitta Böckeler"
 last_updated: 2026-04-11
@@ -1,5 +1,7 @@
 ---
 title: "面向编码智能体用户的 Harness 工程"
 type: concept
 confidence: medium
 source: "https://martinfowler.com/articles/harness-engineering.html"
 author: "Birgitta Böckeler"
 published: 2026-04-02
@@ -1,5 +1,7 @@
 ---
 title: "Harness 工程"
 type: concept
 confidence: medium
 source: "Externalization in LLM Agents: A Unified Review of Memory, Skills, Protocols and Harness Engineering"
 raw_sources:
  - path: raw/Externalization in LLM Agents A Unified Review of Memory, Skills, Protocols and Harness Engineering.md
@@ -1,5 +1,7 @@
 ---
 title: "LLM Wiki 的实现模式"
 type: concept
 confidence: medium
 source: "nashsu/llm_wiki README"
 raw_sources:
  - path: raw/LLM Wiki - Persistent Personal Knowledge Bases.md
@@ -1,5 +1,7 @@
 ---
 title: "LLM Wiki：构建持久累积型个人知识库"
 type: concept
 confidence: medium
 source: "https://github.com/nashsu/llm_wiki/blob/main/llm-wiki.md"
 author: "nashsu"
 created: 2026-08-02
@@ -1,5 +1,7 @@
 ---
 title: "记忆系统"
 type: concept
 confidence: medium
 source: "Externalization in LLM Agents: A Unified Review of Memory, Skills, Protocols and Harness Engineering"
 raw_sources:
  - path: raw/Externalization in LLM Agents A Unified Review of Memory, Skills, Protocols and Harness Engineering.md
@@ -1,5 +1,7 @@
 ---
 title: "Meta-Harness：模型 Harness 的端到端优化"
 type: concept
 confidence: medium
 source: "https://arxiv.org/html/2603.28052v1"
 author: "Yoonho Lee, Roshen Nair, Qizheng Zhang, Kangwook Lee, Omar Khattab, Chelsea Finn"
 published: 2026
@@ -1,5 +1,7 @@
 ---
 title: "技能系统"
 type: concept
 confidence: medium
 source: "Externalization in LLM Agents: A Unified Review of Memory, Skills, Protocols and Harness Engineering"
 raw_sources:
  - path: raw/Externalization in LLM Agents A Unified Review of Memory, Skills, Protocols and Harness Engineering.md
@@ -1,5 +1,7 @@
 ---
 title: "LLM Wiki 桌面实现与外部接入"
 type: practice
 confidence: medium
 source: "nashsu/llm_wiki README_CN; nashsu/llm_wiki mcp-server README"
 raw_sources:
  - path: raw/llm_wiki/README_CN.md
@@ -1,5 +1,7 @@
 ---
 title: "LangChain Harness 工程实践"
 type: practice
 confidence: medium
 source: "https://blog.langchain.com/improving-deep-agents-with-harness-engineering/"
 author: "LangChain"
 published: 2026-02-18
@@ -1,5 +1,7 @@
 ---
 title: "长运行应用的 Harness 设计"
 type: practice
 confidence: medium
 source: "Harness design for long-running application development"
 raw_sources:
  - path: raw/Harness design for long-running application development.md
@@ -1,5 +1,7 @@
 ---
 title: "Managed Agents：将大脑与手分离"
 type: practice
 confidence: medium
 source: "https://www.anthropic.com/engineering/managed-agents"
 author: "Anthropic"
 published: 2026
@@ -1,5 +1,7 @@
 ---
 title: "Mitchell Hashimoto 的 AI 采用之旅"
 type: practice
 confidence: medium
 source: "https://mitchellh.com/writing/my-ai-adoption-journey"
 author: "Mitchell Hashimoto"
 published: 2026-02-05
@@ -1,5 +1,7 @@
 ---
 title: "Obsidian CLI 作为本地 Wiki 维护工具"
 type: practice
 confidence: medium
 source: "https://obsidian.md/zh/help/cli"
 raw_sources:
  - path: raw/Obsidian CLI - Obsidian 中文帮助 (2026_8_2 16：19：05).html
@@ -1,5 +1,7 @@
 ---
 title: "OpenAI Codex Harness 工程"
 type: practice
 confidence: medium
 source: "工程技术：在智能体优先的世界中利用 Codex"
 raw_sources:
  - path: raw/工程技术：在智能体优先的世界中利用 Codex.md
@@ -1,12 +1,15 @@
 ---
 title: "PaddleOCR OCR 工具链使用指南"
 type: practice
 confidence: medium
 source: "PaddleOCR 官方 API 文档 + 2026-08-07 本地实测（wikillm/scripts/paddleocr_v2_ocr.py）"
 tags:
  - "PaddleOCR"
  - "OCR"
  - "工具链"
-sources:
+raw_sources:
  - path: "raw/技术/paddleocr/paddleocr-api-配额与错误码-摘编.md"
 last_updated: 2026-08-07
 ---
 # PaddleOCR OCR 工具链使用指南
@@ -1,6 +1,6 @@
 ---
 title: "利弗莫尔交易纪律"
-type: practices
+type: practice
 last_updated: 2026-08-07
 tags:
  - "交易心理"
@@ -1,5 +1,7 @@
 ---
 title: "用通俗易懂的方式理解 Harness 工程"
 type: query
 confidence: medium
 source: "WikiLLM Q&A"
 date: 2026-04-12
 tags: