chore: save workspace changes before pull

2026-04-20 14:31:39 +08:00
parent 08da9a4d38
commit 9b5a9a9902
47 changed files with 1359 additions and 3033 deletions
--- a/wiki/sources/accounts-payable-agent.md
+++ b/wiki/sources/accounts-payable-agent.md
@@ -0,0 +1,39 @@
+---
+title: "Accounts Payable Agent"
+type: source
+tags: [agent, finance, payments, automation]
+date: 2026-04-20
+last_updated: 2026-04-20
+source_file: raw/Agent/agency-agents/specialized/accounts-payable-agent.md
+---
+
+## Source File
+- [[raw/Agent/agency-agents/specialized/accounts-payable-agent.md]]
+
+## Summary
+- Accounts Payable Agent 是 The Agency 体系中的支付与应付账款专家，负责处理供应商发票、承包商付款与周期性账单，并可在 ACH、Wire、Crypto、Stablecoin 与 Payment API 等多种 payment rail 间自动路由。
+- 该智能体的核心原则是幂等性、严格验证、完整审计与不重复付款，任何超过授权阈值的付款都会被升级处理。
+- 它强调与其他智能体协作，在收到付款请求后完成验证、执行、记录并通知请求方。
+
+## Key Claims
+- 付款前必须检查同一发票是否已经支付，避免重复付款
+- 50 美元以上的付款必须先验证收款方地址或账户
+- 任何付款都需要记录发票编号、金额、rail、时间戳和状态
+- 若某条 payment rail 失败，应尝试下一条可用 rail，再决定是否升级
+- 超出授权限额的付款必须交由人工审批
+
+## Key Quotes
+> "Idempotency first" — 付款安全的首要规则
+> "Never pay twice" — 防止重复付款的核心约束
+> "Audit everything" — 每笔付款都必须留下完整审计轨迹
+
+## Connections
+- [[The Agency]] — 该智能体所属的项目框架
+- [[Idempotent Operation]] — 重复执行也不应产生重复付款
+- [[Audit Trail]] — 每笔付款必须可追踪、可审计
+- [[Payment Rail]] — 选择最优支付通道完成结算
+- [[Spend Limit]] — 超过授权阈值时必须升级审批
+- [[Labor Law Compliance]] — 付款协作场景中可能涉及合规约束
+
+## Contradictions
+- 无明显冲突
--- a/wiki/sources/agentic-identity-trust.md
+++ b/wiki/sources/agentic-identity-trust.md
@@ -0,0 +1,48 @@
+---
+title: "Agentic Identity & Trust Architect"
+type: source
+tags: [agent, the-agency, identity, trust, security, zero-trust, audit]
+date: 2026-04-20
+last_updated: 2026-04-20
+---
+
+## Source File
+- [[raw/Agent/agency-agents/specialized/agentic-identity-trust.md]]
+
+## Summary
+- Agentic Identity & Trust Architect is The Agency's zero-trust specialist for autonomous agents, focused on cryptographic identity, delegated authorization, trust scoring, and tamper-evident evidence.
+- The role separates agent identity from authorization and insists that every consequential action be backed by verifiable proof, not self-reported claims.
+- It complements [[Identity Graph Operator]], which resolves entity identity, by providing the agent-side identity and trust layer.
+
+## Key Claims
+- Agents must prove who they are with cryptographic identity checks; self-reported identity is not enough.
+- Authorization must be scoped, revocable, and verifiable through delegation chains.
+- Trust should start at zero and only increase through verifiable outcomes, fresh credentials, and intact evidence chains.
+- Evidence records must be append-only and tamper-evident; if evidence cannot be written, the action should not proceed.
+- Algorithm agility and post-quantum migration readiness should be designed in from the start.
+
+## Key Quotes
+> "Never trust self-reported identity." — zero-trust rule for agent networks
+
+> "If evidence cannot be written, the action should not proceed." — fail-closed authorization rule
+
+## Key Concepts
+- [[Identity Governance]]
+- [[Audit Trail]]
+- [[Zero Trust Access]]
+- [[Identity Graph Operator]]
+- [[The Agency]]
+
+## Key Entities
+- [[Agentic Identity & Trust Architect]]
+- [[Identity Graph Operator]]
+- [[The Agency]]
+
+## Connections
+- [[The Agency]] ← contains ← [[Agentic Identity & Trust Architect]]
+- [[Agentic Identity & Trust Architect]] ← complements ← [[Identity Graph Operator]]
+- [[Identity Governance]] ← informed_by ← [[Agentic Identity & Trust Architect]]
+- [[Audit Trail]] ← constrains ← [[Agentic Identity & Trust Architect]]
+
+## Contradictions
+- None noted
--- a/wiki/sources/government-digital-presales-consultant.md
+++ b/wiki/sources/government-digital-presales-consultant.md
@@ -0,0 +1,56 @@
+---
+title: "Government Digital Presales Consultant"
+type: source
+tags: [agent, the-agency, presales, government-it, compliance]
+date: 2026-04-20
+---
+
+## Source File
+- [[raw/Agent/agency-agents/specialized/government-digital-presales-consultant.md]]
+
+## Summary
+- 核心主题：面向中国政府数字化转型市场（ToG）的售前专家角色，覆盖政策解读、方案设计、招投标、POC 验证与合规适配
+- 问题域：政务项目机会识别难、招投标流程复杂、合规要求严格、技术方案需要同时满足业务与监管约束
+- 方法/机制：以政策信号驱动机会发现，以业务场景驱动方案设计，以 POC 和答标材料完成赢单闭环
+- 结论/价值：将政府项目售前从“拼概念”转为“拼合规、拼方案、拼交付可行性”，帮助团队更稳定地推进项目成交
+
+## Key Claims
+- 政务数字化项目要从国家/省市政策和预算信号中识别机会，而不是只等招标公告
+- Dengbao、Miping 与 Xinchuang 不是加分项，而是政府系统常见的硬约束
+- 售前方案必须围绕业务场景表达价值，避免只讲架构和技术名词
+- 评分高低往往取决于对招标文件条款、评分规则与资格条件的精准响应
+- POC 的作用是证明关键能力，不是无限期免费试做
+- 采购、技术、业务和领导层关注点不同，表达方式必须分层
+
+## Key Quotes
+> "Dengbao、Miping 和 Xinchuang are mandatory, not bonus points." — 合规底线
+
+> "The client cares about service outcomes, not microservices architecture." — 方案表达原则
+
+## Key Concepts
+- [[Technical Discovery]]
+- [[Solution Architecture (SA)]]
+- [[Proof of Concept (POC)]]
+- [[Digital Government]]
+- [[Smart City]]
+- [[Government Procurement]]
+- [[Dengbao]]
+- [[Miping]]
+- [[Xinchuang]]
+- [[Compliance Enforcement]]
+
+## Key Entities
+- [[The Agency]]
+
+## Connections
+- [[Government Digital Presales Consultant]] ← role within ← [[The Agency]]
+- [[Government Digital Presales Consultant]] ← applies ← [[Technical Discovery]]
+- [[Government Digital Presales Consultant]] ← shapes ← [[Solution Architecture (SA)]]
+- [[Government Digital Presales Consultant]] ← validates via ← [[Proof of Concept (POC)]]
+- [[Government Digital Presales Consultant]] ← constrained by ← [[Government Procurement]]
+- [[Government Digital Presales Consultant]] ← constrained by ← [[Dengbao]]
+- [[Government Digital Presales Consultant]] ← constrained by ← [[Miping]]
+- [[Government Digital Presales Consultant]] ← constrained by ← [[Xinchuang]]
+
+## Contradictions
+- 无明显冲突
--- a/wiki/sources/identity-graph-operator.md
+++ b/wiki/sources/identity-graph-operator.md
@@ -0,0 +1,45 @@
+---
+title: "Identity Graph Operator"
+type: source
+tags: [agent, the-agency, identity, graph, entity-resolution]
+date: 2026-04-20
+last_updated: 2026-04-20
+---
+
+## Source File
+- [[raw/Agent/agency-agents/specialized/identity-graph-operator.md]]
+
+## Summary
+- 核心主题：Identity Graph Operator 负责多智能体系统中的共享身份图与实体归一化，确保不同 agent 面对同一现实世界实体时得到同一 canonical identity
+- 问题域：重复记录、冲突写入、跨 agent 身份不一致、合并/拆分决策难以追踪
+- 方法/机制：通过 blocking、scoring、clustering 和乐观锁在单一引擎内执行身份解析、合并提案与冲突检测
+- 结论/价值：为多智能体系统提供确定性、可审计、可回滚的身份层，降低重复创建与级联错误风险
+
+## Key Claims
+- 同一输入必须得到同一 entity_id，determinism 是身份解析的首要原则
+- 任何 merge/split/update 都应经过单一引擎处理，并保留事件历史与回滚能力
+- 解析结果应返回 confidence、canonical_data、version 等可审计字段
+- 多 agent 协作时优先提案而非直接突变，以便人工或其他 agent 审核
+- 共享身份层需要 tenant isolation，并默认对 PII 进行脱敏
+
+## Key Quotes
+> "Same input, same output. Two agents resolving the same record must get the same entity_id." — 设计原则
+
+## Key Concepts
+- [[AI代理（Agent）]]：Identity Graph Operator 是支撑多智能体协作可靠性的基础设施
+- [[Audit Trail]]：所有合并、拆分、匹配都应有理由和置信度记录
+- [[Identity Governance]]：共享身份层的治理框架，强调 canonical identity 与权限边界
+- [[Multi-Agent-System-Reliability]]：身份层是多智能体系统可靠性的前提之一
+
+## Key Entities
+- [[The Agency]]：开源 AI 智能体集合项目，Identity Graph Operator 所属项目语境
+- [[Identity Graph Operator]]：该源文档对应的专门智能体角色
+
+## Connections
+- [[The Agency]] ← contains ← [[Identity Graph Operator]]
+- [[Identity Graph Operator]] ← supports ← [[AI代理（Agent）]]
+- [[Identity Graph Operator]] ← constrains_by ← [[Audit Trail]]
+- [[Identity Graph Operator]] ← relies_on ← [[Identity Governance]]
+
+## Contradictions
+- 未发现明显冲突
--- a/wiki/sources/project-management-experiment-tracker.md
+++ b/wiki/sources/project-management-experiment-tracker.md
@@ -0,0 +1,41 @@
+---
+title: "Experiment Tracker"
+type: source
+tags: [agent, project-management]
+date: 2026-04-20
+source_file: raw/Agent/agency-agents/project-management/project-management-experiment-tracker.md
+---
+
+## Source File
+- [[raw/Agent/agency-agents/project-management/project-management-experiment-tracker.md]]
+
+## Summary
+- 核心主题：Experiment Tracker 是一位专注于实验设计、执行跟踪和数据驱动决策的项目管理智能体
+- 问题域：A/B 测试、特性实验、假设验证与实验组合管理
+- 方法/机制：明确假设、统计显著性、样本量计算、随机分配、风险监控与回滚
+- 结论/价值：通过严谨的实验方法确保产品决策建立在可验证数据之上
+
+## Key Claims
+- 设计实验前应先完成样本量与统计功效分析，默认追求 95% 统计置信度
+- 实验必须保持随机分配、避免抽样偏差，并在多方案比较时进行多重比较修正
+- 不应在没有明确早停规则的情况下提前终止实验
+- 实验安全需要监控用户体验退化、隐私合规和回滚机制
+
+## Key Quotes
+> "Always calculate proper sample sizes before experiment launch" — statistics and rigor rule
+
+> "Never stop experiments early without proper early stopping rules" — experiment integrity rule
+
+## Connections
+- [[The Agency]] — 所属智能体集合项目
+- [[Senior Project Manager]] — 同属项目管理语境，但更偏向任务拆解与交付执行
+- [[Project Shepherd]] — 同属项目管理语境，但更偏向跨职能协调与交付治理
+- [[Studio Operations]] — 同属项目管理语境，但更偏向运营流程和资源协调
+- [[A/B Testing]] — 核心实验类型之一
+- [[Hypothesis Testing]] — 实验设计的统计基础
+- [[Statistical Significance]] — 决策判定依据
+- [[Power Analysis]] — 样本量与检验功效的计算方法
+- [[Randomization]] — 控制偏差的实验设计原则
+
+## Contradictions
+- 与直觉驱动决策存在冲突：该智能体强调先验证、后推广，反对未经验证的主观拍板。
--- a/wiki/sources/recruitment-specialist.md
+++ b/wiki/sources/recruitment-specialist.md
@@ -0,0 +1,51 @@
+---
+title: "Recruitment Specialist"
+type: source
+tags: [agent, the-agency, recruitment, hr, compliance]
+date: 2026-04-20
+last_updated: 2026-04-20
+---
+
+## Source File
+- [[raw/Agent/agency-agents/specialized/recruitment-specialist.md]]
+
+## Summary
+- 核心主题：Recruitment Specialist 是 The Agency 体系中的招聘运营与人才获取专家，覆盖中国主流招聘平台、人才评估、面试设计、入职管理和劳动法合规
+- 问题域：招聘渠道效率、JD 质量、候选人体验、用工合规、offer/入职流程
+- 方法/机制：通过多渠道投放、结构化面试、数据驱动筛选、雇主品牌内容和 ATS 流程实现端到端招聘运营
+- 结论/价值：为企业提供可量化、可审计、合规的全流程招聘系统，兼顾效率、候选人体验与法律风险控制
+
+## Key Claims
+- Boss Zhipin、Lagou、Liepin、Zhaopin、51job、Maimai 和 LinkedIn China 各有不同的岗位适配与 ROI 侧重点
+- JD 应区分硬性要求与软性偏好，并从候选人视角强调文化、成长和福利
+- 所有招聘决策都应以数据支持，避免凭直觉选人
+- 候选人个人信息、背景调查和不竞争义务筛查必须符合法律与授权要求
+- 初筛反馈应在 48 小时内完成，以维护候选人体验和雇主口碑
+
+## Key Concepts
+- [[Recruitment Operations]]：多渠道招聘投放、预算分配和漏斗优化
+- [[Talent Assessment]]：通过结构化与行为面试评估胜任力的流程
+- [[Labor Law Compliance]]：招聘、签约、背景调查与用工全过程合规
+- [[Candidate Experience]]：候选人从投递到 offer 的体验设计
+- [[Employer Brand]]：通过内容与口碑构建雇主吸引力
+- [[ATS]]：用系统化流程管理招聘漏斗和候选人数据
+- [[Onboarding]]：从 offer 到入职与试用期管理的标准流程
+
+## Key Entities
+- [[The Agency]]：Recruitment Specialist 所属项目框架
+- [[Boss Zhipin]]：主打直聊与曝光优化的招聘平台
+- [[Liepin]]：适合中高阶岗位的猎头平台
+- [[Maimai]]：适合雇主品牌与被动候选人触达的职业社交平台
+
+## Connections
+- [[The Agency]] ← contains ← [[Recruitment Specialist]]
+- [[Recruitment Specialist]] ← supports ← [[Recruitment Operations]]
+- [[Recruitment Specialist]] ← applies ← [[Talent Assessment]]
+- [[Recruitment Specialist]] ← constrained_by ← [[Labor Law Compliance]]
+- [[Recruitment Specialist]] ← improves ← [[Candidate Experience]]
+- [[Recruitment Specialist]] ← builds ← [[Employer Brand]]
+- [[Recruitment Specialist]] ← operationalized_by ← [[ATS]]
+- [[Recruitment Specialist]] ← extends_to ← [[Onboarding]]
+
+## Contradictions
+- 无明显冲突
--- a/wiki/sources/specialized-civil-engineer.md
+++ b/wiki/sources/specialized-civil-engineer.md
@@ -0,0 +1,39 @@
+---
+title: "Civil Engineer"
+type: source
+tags: [agent, the-agency, engineering, civil]
+date: 2026-04-20
+---
+
+## Source File
+- [[raw/Agent/agency-agents/specialized/specialized-civil-engineer.md]]
+
+## Summary
+- 这份智能体规范定义了 Civil Engineer，一位严谨的土木/结构工程专家，强调在多国标准下输出安全、经济、可施工的设计
+- 主题覆盖结构分析、岩土评估、施工图与技术说明、建筑规范合规，以及多标准项目中的冲突协调
+- 核心价值在于：每个设计都必须明确治理规范、版本、荷载组合与关键假设，并同时检查 ULS 与 SLS
+- 适用场景包括跨 Eurocode、ACI、AISC、AS/NZS、CSA、GB、IS、AIJ 等标准的国际项目
+
+## Key Claims
+- Civil Engineer 必须同时处理结构安全、经济性和可施工性，而不是只追求理论强度
+- 每个计算包都应明确治理规范、版本、国家附录/地方修订与关键假设
+- 设计必须同时验证 ULS 和 SLS，不能只做强度校核
+- 岩土参数不能凭空假设，必须基于勘察报告或清晰的书面假设
+- 多标准项目需要为每个构件/子系统明确适用规范，并记录冲突解决策略
+
+## Key Quotes
+> "Always state the governing code edition and national annex at the start of every calculation package." — 规范与版本要求
+
+> "Never assume soil parameters without a ground investigation report or clear stated assumptions." — 岩土严谨性要求
+
+> "The section passes strength (ULS) but deflection (SLS) governs." — 结构设计中强度与使用性并重
+
+## Connections
+- [[Civil Engineer]] — 该 source 对应的实体页面
+- [[The Agency]] — The Agency 体系中的专业化工程智能体之一
+- [[Multi-Agent Team]] — 体现专职角色 + 共享上下文的多智能体协作模式
+- [[Technical Discovery]] — 项目前期必须先明确规范、荷载、边界条件与决策标准
+- [[Global Standards Coverage]] — 覆盖 Eurocode、ACI、AISC、AS/NZS、GB、IS、AIJ 等多规范体系
+
+## Contradictions
+- 无明显冲突
--- a/wiki/sources/specialized-document-generator.md
+++ b/wiki/sources/specialized-document-generator.md
@@ -0,0 +1,47 @@
+---
+title: "Document Generator"
+type: source
+tags: [agent, the-agency, document-generation, office-suite]
+date: 2026-04-20
+---
+
+## Source File
+- [[raw/Agent/agency-agents/specialized/specialized-document-generator.md]]
+
+## Summary
+- 核心主题：Document Generator 专家智能体负责以代码方式生成 PDF、PPTX、XLSX 和 DOCX 等专业文档
+- 问题域：从零手工排版、格式不一致、难以复用模板、数据可视化质量不稳定
+- 方法/机制：使用 Python/Node.js 文档生成库，结合模板、样式和数据驱动流程输出成品文件
+- 结论/价值：将文档生产变成可重复、可自动化的工程流程，而不是一次性手工设计
+
+## Key Claims
+- Document Generator 擅长使用代码化方式生成专业文档
+- PDF 可用 reportlab、weasyprint、fpdf2、puppeteer 等工具生成
+- PPTX 可用 python-pptx 或 pptxgenjs 生成
+- XLSX 可用 openpyxl、xlsxwriter、exceljs 生成
+- DOCX 可用 python-docx 或 docx 生成
+- 文档必须使用样式与主题，而不是硬编码字体和字号
+- 输出应兼顾品牌一致性、可访问性与模板复用
+- 生成前应确认受众与目的，并交付脚本与成品文件
+
+## Key Quotes
+> "Generate professional documents using the right tool for each format." — 文档格式选择原则
+
+> "Use proper styles — Never hardcode fonts/sizes; use document styles and themes." — 关键规则
+
+## Key Concepts
+- [[Claude Skills]]：将重复性的文档生成流程封装为可复用 SOP
+- [[Prompt Engineering]]：需要清晰的输入、上下文和输出格式
+- [[Document Generation]]：代码化文档生产与模板化输出的工作流
+
+## Key Entities
+- [[The Agency]]：所属项目
+- [[Document Generator]]：对应的 AI Agent 实体
+
+## Connections
+- [[The Agency]] ← contains ← [[Document Generator]]
+- [[Claude Skills]] ← enables ← [[Document Generator]]
+- [[Prompt Engineering]] ← supports ← [[Document Generator]]
+
+## Contradictions
+- 无明显冲突