claude-skills

Create and iteratively refine parametric 3D CAD models through progressive refinement with inspectable feedback. Use when the task involves OpenSCAD, build123d, STL/STEP/3MF output, fixture/enclosure/adapter design, or debugging shape accuracy from renders or screenshots. Covers both generation and editing as a unified workflow.

Claudeception is a continuous learning system that extracts reusable knowledge from work sessions. Triggers: (1) /claudeception command to review session learnings, (2) "save this as a skill" or "extract a skill from this", (3) "what did we learn?", (4) After any task involving non-obvious debugging, workarounds, or trial-and-error discovery. Creates new Claude Code skills when valuable, reusable knowledge is identified.

AI-assisted distributed systems verification with TLA+ formal specification. Use when verifying distributed system correctness, discovering race conditions, writing TLA+ specifications from codebases, performing model checking on consensus/replication protocols, or setting up Lamport Agent workflows. Triggers on: "verify distributed system", "formal verification", "TLA+ spec", "model checking", "find race condition", "prove correctness", "CRAQ", "Paxos verification", "Raft verification", "chain replication verification", "distributed consensus verification".

Expert tutor and reference for 具身智能 / embodied intelligence and robot learning, grounded in 《具身智能导论》— the 17-chapter Chinese textbook translated from UC Berkeley CS294-291 "Robots That Learn". Use this skill whenever the user asks about embodied AI / robot learning topics the book covers: 中枢模式发生器(CPG)、 生物运动力学与步态、李群 SO(3)/SE(3) 机器人运动学与动力学、扩散模型/扩散策略 (DDPM, score matching, flow matching, diffusion policy)、灵巧手与欠驱动协同、 触觉感知(GelSight)、运动控制(LQR/MPC/阻抗控制)、卡尔曼滤波与预测控制、视频世界模型、 强化学习(MDP/SAC/PPO/actor-critic)、行为克隆与动作分块、视觉模仿学习(UMI/跨具身)、 足式运动(RMA)、导航(GOAT/可穿越性)、Sim-to-Real 灵巧操作、VLA 与 π0.5 长程规划。 Trigger it for conceptual questions, math derivations, paper explanations, PyTorch implementation help, study/teaching, or "解释一下/推导/这门课/具身智能" requests in this domain — even when the user doesn't name the book. Prefer this skill over answering from memory, because it gives grounded, citable, notation- consistent answers and links to the exact chapter.

Automates publishing novel chapters to 番茄小说 (fanqienovel.com) via Playwright browser automation. Use when the user wants to publish, batch-upload, or retry uploading novel chapters from a local directory to 番茄小说. Handles login state, ProseMirror editor injection, popup dismissal (错别字检测/仅基础检测/AI声明/确认发布), and auto-archives successfully published files. Triggers on requests like "发布章节到番茄", "上传小说", "batch publish chapters", or retrying failed chapters.

Use when reviewing HarmonyOS (ArkTS/ArkUI) code, when file extensions are .ets/.ts in HarmonyOS context, or when checking ArkUI component implementations, async/await patterns, or @State/@Prop/@Provide usage

HarmonyOS/OpenHarmony Node-API (NAPI) domain expertise for developing native C/C++ modules that bridge ArkTS/JS with native code. Use when: (1) Creating or modifying NAPI native modules in OpenHarmony projects, (2) Implementing high-performance native-codec/image-processing/thumbnail NAPI interfaces, (3) Designing NAPI module registration, async work, thread-safe callbacks, or Promise-based native APIs, (4) Debugging NAPI lifecycle, memory, or cross-thread issues, (5) Applying NAPI performance patterns (ArrayBuffer zero-copy, Sendable types, QoS scheduling), (6) Working with multimedia_media_library or any OHOS service-layer NAPI code.

HarmonyOS/OpenHarmony System Ability Manager (SAMgr) daemon domain expertise. Covers SA architecture, lifecycle management, IPC/RPC communication patterns, service registration/discovery, source code structure, and API reference. Use when: (1) Working with OpenHarmony SAMgr / systemabilitymgr_samgr code, (2) Questions about SA lifecycle, daemon startup, or boot sequence, (3) Implementing or debugging IPC/RPC Proxy/Stub patterns for system services, (4) Understanding SA registration (AddSystemAbility, GetSystemAbility, LoadSystemAbility), (5) Analyzing SAMgr crash recovery, on-demand loading, or distributed SA routing, (6) Navigating the systemabilitymgr subsystem source tree.

Solve any problem in the philosophical family of the knapsack problem (背包问题) using the unified state-design framework from 崔添翼《背包问题九讲》. Use whenever a problem reduces to "choose a subset/multiset of items under a capacity-like cost to optimize a value-like objective" — including 0/1 pack, unbounded/complete pack, bounded/multiple pack, mixed pack, 2-D cost pack, grouped pack, dependency pack (主件/附件, tree-shaped), generalized items (泛化物品), subset-sum / can-it-fill-exactly, counting feasible or optimal schemes, recovering the chosen subset, lexicographically smallest plan, and Kth-best value. Trigger words — 背包/knapsack, subset sum, "choose items with weight/cost ≤ W to maximize value", "ways to make change", "partition into K groups", "limited copies", "items depend on each other", "fill exactly", coin change counting, NOIP-style 金明的预算方案 / 采药 / 装箱问题.

Electron application reverse engineering — ASAR unpacking, webpack bundle analysis, auth/payment logic discovery, IPC channel mapping, and client-side authorization bypass analysis. Use when the user needs to: reverse engineer an Electron app (.deb/.dmg/.exe), analyze auth/subscription/license logic in Electron apps, map IPC channels between main and renderer processes, identify client-side enforcement vulnerabilities (CWE-602), or perform security audit of Electron-based desktop applications for CTF or research.

AI-assisted Rust development with code contracts, lightweight spec-driven development, and performance optimization. Use when writing, reviewing, or optimizing Rust code -- especially for distributed systems, consensus protocols, or any correctness-critical Rust project. Triggers on Rust development tasks, performance tuning requests, "write Rust code with contracts", "optimize Rust performance", "spec-driven Rust development", or AI-assisted Rust workflows modeled on 100K+ line production-grade Rust projects.

This skill should be used when the user asks to "write C code with sp.h", "use spclib", "sp.h API", "single-header C library", "modern C programming", "sp_str_t usage", "sp_alloc memory", "sp_log formatting", "dynamic array in C", "hash table in C", "cross-platform C code", or when working with the sp.h single-header C standard library replacement.

End-to-end workflow for translating an English course or book (lecture notes, slides, and reading papers) into a structured Chinese Markdown book published via GitHub Pages (Jekyll). Use this whenever the user wants to translate a course/textbook/lecture series/paper collection into Chinese as a multi-chapter Markdown site or book — phrases like "把这门课翻译成中文教程", "translate this course to a Chinese book", "做成中文电子书发到 GitHub Pages", "翻译 lecture notes 输出学习手册". ALSO use this skill whenever math/LaTeX formulas render incorrectly on a Jekyll / GitHub Pages site (kramdown + MathJax) — broken subscripts, formulas showing as raw text, equations turning into tables, italic leaking into math — even if no translation is involved, because the math-rendering reference here is the authoritative fix. Covers PDF ingestion, parallel per-chapter translation, terminology consistency, and the GitHub Pages math-rendering minefield.

Up-to-date Zig programming language patterns for version 0.16.0. Use when writing, reviewing, or debugging Zig code, working with build.zig and build.zig.zon files, or using comptime metaprogramming. Critical for avoiding outdated patterns from training data - especially build system APIs (root_module instead of root_source_file), I/O APIs (buffered writer pattern), container initialization (.empty/.init), allocator selection (DebugAllocator), and removed language features (@Type, @cImport, async/await, usingnamespace).

MobAI Pro 会员系统逆向工程全流程。适用于 Go 64-bit PE 分析、license/quota 系统 patch、NSIS→7z SFX 重打包、以及 Go 二进制 embed.FS 前端提取。需要 IDA Pro MCP 连接。