@wangjiaocheng

knowledge-explainer atomizes concepts into irreducible units, orders them by dependency, and reconstructs complete explanations via formulas, derivations, intuition, and misconception analysis. targets education and knowledge system building.

academic-thesis-workflow automates four-step scholarly paper generation from topic to complete thesis with embedded feasibility assessment and quality checkpoints. covers humanities and stem fields via field-adaptive argument structures and embedded reference samples.

universal-agent provides a three-mode workflow (standalone, bridge, simulation) for executing natural language tasks via dynamic code generation and shell execution. includes safety checks and error retry, but lacks detailed step-by-step procedures and comprehensive failure mode documentation.

ai-dev-workflow structures ai-assisted programming in three ordered steps (requirement to feature description, description plus tech stack to method calls, calls to full project code) using embedded prompts and reference samples. intended for building complete systems from zero.

cda-code-lab generates executable python simulation code for causal dynamics architecture components including cdablock pipelines, pinn mechanisms, hamiltonian projections, and symplectic integrators. code reads json from data synthesizer and targets pytorch-based prototyping of phase 1-2 architecture layers.

universal-primitives frames llm capability architecture around two primitives (file crud, script execution) and their combinations. covers agent design philosophy, hardware control pathways, and environment self-bootstrapping strategies for incomplete tool environments.

c2c-platform-skill-system is a 310-unit taxonomy across 10 operational clusters (market intelligence, content, growth, analytics, knowledge, compliance, supply-demand, operations, qualification, infrastructure). routing to reference files and pipelines replaces actionable procedure; execution logic exists only as loading rules without decision trees or recovery paths.

web-novel-writing-reference provides a taxonomy of 62 writing tasks across 8 domains (worldbuilding, outline, character, plot, scene, pacing, revision, business) for chinese web fiction, organized as reference material for the UTOS execution framework. core value is structural classification, not actionable procedure.

singlefile-output-reference catalogs single-file output types across html, python, shell, and sql with decision trees and exemplar references. execution requires universal-task-os integration; without it, degrades to read-only mode.

universal-business-skill-system defines a 10-cluster taxonomy of 286 business capability units across industries. lacks concrete procedures, decision logic, and failure modes; operates as a catalog referencing external files rather than executable skill definition.

smart-hardware-reference documents a 9-domain task taxonomy for embedded product development (68 task types across requirements through service lifecycle), positioned as reference-only metadata for the universal-task-os framework rather than executable procedures.

capability-pipeline-os proposes a unitized task decomposition framework with six meta-operation clusters (sense, cognize, act, organize, interact, guard). the framework is conceptually ambitious but lacks concrete procedural steps, executable examples, and failure mode guidance.

pharma-doc-reference is a metadata repository of 93 pharmaceutical document types organized across 11 domains, intended as a payload library for universal-task-os rather than a standalone execution skill. it provides document catalogs, content requirement checklists, and exemplar templates but delegates all actual workflow execution to utos.

metabolic-healing-skill-system claims 267 atomic units across 10 clusters for non-pharmacological chronic disease reversal, but lacks procedural detail, step-by-step execution paths, and testable decision logic. references remain unwritten.

innovation-os frames creative problem-solving around four pattern-processing modes (direct application, improvement, migration, construction) and ten meta-frameworks. actual procedures absent from this document, deferred to external reference.

pharma-skill-system outlines a 10-cluster taxonomy for pharmaceutical workflows covering 334 units across intelligence, content, compliance, and supply chain domains. it references external files for unit definitions but provides no executable procedures, decision logic, or outcome signals.

minimal-agent wraps system-controller scripts for windows automation, process control, hardware adjustment, gui manipulation, serial communication and iot interaction. dual-mode architecture (v1 text, v2 function) promises flexibility but lacks procedural clarity.

universal-task-os claims a three-axis decomposition for task handling (execution, content, innovation) but provides no concrete procedures, examples, or decision logic. references external files and meta-structures without actionable steps.

wealth-manager frames a four-stage wealth lifecycle (accumulation, growth, preservation, succession) with references to trading laws, risk discipline, and life strategy algorithms. execution details are deferred to external markdown files and interactive calculators.

cda provides reference documentation for a proposed causal dynamics architecture as transformer alternative, structured around entity-state graphs and hamiltonian constraints. skill offers deep-dive capability into five-layer compute stack but lacks actionable procedures or decision flows.

omniscient claims universal capability across cognition, code execution, and system control via three layers plus an orchestration engine. the architecture diagram and module list are comprehensive, but the skill lacks concrete trigger phrases, actionable step-by-step procedures, and edge case handling.

linux-omniscient claims to unify cognition, code execution, and system control across three layers plus an orchestration engine, but lacks concrete trigger phrases, actionable step-by-step procedures, and tested workflows. the document reads as architectural vision rather than a validated, deployable skill.

双语对话技能——用户输入和 AI 回复均以四层双语对照格式展示（中文→拼音→英文→IPA），支持可选层级省略和多语言扩展。 Bilingual conversation skill — both user input and AI replies are displayed in a four-layer bil...

CDA 因果数据合成器——根据领域描述自动生成 CDA 架构可消费的因果数据集。 生成 Entity-State Graph（JSON 格式）、CausalEdge 列表、带时间戳的观测轨迹、哈密顿量约束参数。 覆盖热力学、力学、流体、多域耦合等物理域，支持生成训练数据、验证数据、基准测试数据。 当用户需要为 C...

代理记账行业十大功能集群技能体系。基于"任务为中心，AI Pipeline驱动"思想，覆盖市场情报、内容获客、增长引擎、经营分析、财税知识、合规风控、客户全周期、服务流程、资质准入、工具与供应链十大业务流。专注代理记账、工商注册、税务申报、汇算清缴、税务筹划、财税咨询等代理记账机构全业务场景。触发词：代理记账、代...

运用四种AI模式处理能力（模式直用、模式改进、模式迁移、模式构建）为任意学科领域生成符合学术要求的创新选题。模式改进内置10种创新元框架（第一性原理、逆向思维、辩证综合、随机性驱动、涌现生成、演化迭代、系统动力学、约束驱动、故事叙述、游戏化），模式直用的基元可运用所有模式创新且基元重组本身也是创新方式，模式迁移和...

通用内容构成方法论技能，提供清单法和样本法两种范式来生成任意类型的结构化成品。当用户要求生成文档、方案、代码、报告、演讲稿、商业计划书等结构化内容，且明确提到清单法、样本法、组件、样本、模板、模仿等关键词，或要求生成结构化成品但未指定生产方式时触发。此技能不替代领域专业技能，而是在用户选择后指导AI执行对应的构成...

Control Windows desktop software, hardware, and IoT devices. This skill should be used when the user wants to open/close/resize windows, start/stop/monitor p...

Linux系统控制能力包；当用户需要控制桌面窗口、管理进程、调整音量亮度、管理电源网络、与串口设备通信、控制智能家居设备、进行鼠标键盘自动化、截图OCR或视觉识别时使用

认知执行技能 — 整合认知套件与执行框架两大能力层，并配备编排引擎。 认知层：四种思维操作码（直用/改进/迁移/构建）覆盖所有思考任务； 执行层：大语言模型 + 命令执行工具，自动化代码生成与脚本执行； 编排引擎：将所有操作统一为基元，自由组合为任意复杂度的执行链条。 此技能应用于认知与代码执行类任务。

跨平台基础工具集 — 16 个独立可组合的 Python 脚本，替代 Agent 平台缺失的基础工具能力。 文件系统（浏览/搜索/读写/替换/删除）、内容搜索（正则）、网络（搜索/抓取/预览）、 运行时安装、持久化记忆、定时任务、任务管理。 纯 Python 标准库，零外部依赖，跨 Windows/macOS/L...

共享目录消息总线 — 让不同用户/Agent 之间通过文件系统实现聊天对话。 支持单聊、群聊、广播、消息历史查询。 纯 Python 标准库，零外部依赖，跨 Windows/macOS/Linux。 通信基于共享目录（NAS/云同步/网络驱动器），用户自行配置共享路径。