AI & ML

hyqibot Skill

将用户问题中的主题、关键词映射到 CoPaw 官方文档路径与常见源码入口，减少盲目搜索。适用于内置 QA Agent 在回答安装、配置、技能、MCP、多智能体、记忆、CLI 等问题时快速选定要读的文件。

hyqibot Skill

Use this skill when another agent's expertise/context is needed, or when the user explicitly asks to involve another agent. First list agents, then use copaw agents chat for two-way communication with replies. | 当需要其他 agent 的专长/上下文，或用户明确要求调用其他 agent 时使用；先查 agent，再用 copaw agents chat 双向通信（有回复）

lzbovani Bundle

Cria, melhora e revisa prompts profissionais para ChatGPT, Claude, Gemini e outras IAs. Use quando o usuario pedir prompts para texto, codigo, imagem, agentes, estudo, trabalho, automacoes ou qualquer tarefa que precise de prompt otimizado. Antes de gerar o prompt final, deve fazer perguntas criticas para extrair objetivo, contexto, restricoes, publico, formato e criterio de sucesso, salvo quando o usuario pedir explicitamente para gerar direto.

lzbovani Bundle

Gera um handoff completo da sessao atual para iniciar uma nova sessao com continuidade segura. Use quando o usuario pedir resumo da sessao, transferencia de contexto, prompt para proxima sessao, checklist de arquivos necessarios, estado de tarefas, ou retomada de trabalho sem perder historico.

yevhens-hue Skill

Prompt Engineer

yevhens-hue Skill

Multi-Agent Systems & Handoff Architectures using OpenAI Agents SDK and Swarm patterns. Covers @function_tool decorators, Agent & Runner primitives, Handoffs via handoff_description metadata, Manager (Agent-as-Tool) vs Peer Handoff topologies, and Input/Output Guardrails with Tripwires.

yevhens-hue Skill

Signal-Driven Autonomous AI Agent Architecture. Implements event-driven triggers, market/lead signal classification, state machine transitions, and automated multi-step tool execution with audit trails.

yevhens-hue Skill

Production PostgreSQL & pgvector Hybrid Vector Search RAG Architecture. Implements native vector embeddings with HNSW/IVFFlat indexes, full-text TSVECTOR search, in-database Reciprocal Rank Fusion (RRF), Supabase/Prisma ORM integration, and connection pooling.

yevhens-hue Skill

Hierarchical & Sequential Multi-Agent Team Topologies. Covers role specialization (Researcher, Writer, Reviewer, Engineer), shared state passing, task dependency DAGs, error recovery, and consensus voting algorithms for autonomous task resolution.

yevhens-hue Skill

Generative Engine Optimization (GEO) and Answer Engine Optimization (AEO) rules, structured entity grounding (JSON-LD), inverted pyramid content formatting, and SOPs for training AI agents to maximize visibility across Google AI Overviews, Perplexity, and ChatGPT Search.

yevhens-hue Bundle

Structured self-debugging workflow for AI agent failures using capture, diagnosis, contained recovery, and introspection reports.

cnfjlhj Skill

Use when the user asks whether a skill exists, wants help finding or installing one, or asks how to extend agent capabilities with a reusable skill. Start with `npx skills find`; if that fails, use a GitHub-aware fallback before concluding no good match exists.

cnfjlhj Bundle

Use when the user wants multi-agent division of labor for research-led work and the lead should stay on the critical path while 1-2 bounded sidecars handle low-coupling tasks. Do not use this for tiny tasks, fully sequential debugging, or overlapping refactors.

aaronbrethorst Skill

Investigate and fix production errors from Sentry. Use when given a Sentry issue URL, numeric issue ID, or short ID (e.g. VOTERLETTERS-9R). Fetches full issue data via the Sentry MCP, analyzes root cause, scores fix confidence, and applies a TDD red-to-green fix in the Rails app. Triggers on "fix this Sentry issue", "investigate PROJ-XX", any sentry.io URL, or requests to debug a production error from Sentry.

aaronbrethorst Bundle

Guide for the C -fbounds-safety language extension. Covers the language model, pointer annotations, adopting bounds-safety in existing C code, compiler build settings and modes, and runtime debugging of bounds violations.

aaronbrethorst Skill

Use when the user wants to generate UI, prototypes, dashboards, presentations, or other design artifacts via Claude Design (claude.ai/design) and asks for help shaping the prompt — especially when the artifact must reflect project data models, multi-step flows, brand systems, or domain context. Triggers on phrases like "write a Claude Design prompt", "draft a prompt for claude.ai/design", "I want to mock this up in Claude Design".

anbanai Skill

Use when user mentions "初始化", "anban-setup", "第一次使用", "API Key", "密钥", or when MCP tools fail with auth/connection errors suggesting missing ANBAN_API_KEY.

sailingnaturali Skill

Use when multiple sessions work the same repo in parallel — several agent/dev sessions, a long-running review or build that must not be disturbed, or a quick look at another branch or PR. git worktree gives every branch its own directory (own node_modules, own build output) over one shared object store, and git itself enforces that a branch is checked out in at most one place. Covers add/list/remove/prune, per-worktree installs, the rm -rf stale-registration gotcha, and why worktrees and scratch checkouts belong on disk-backed paths, never tmpfs /tmp.

sailingnaturali Skill

Use when creating a repository or upgrading its scaffolding — agent context files, line-ending normalization, format/lint CI, container CVE scanning, tag-triggered publishing, release automation, and generated release notes. Encodes the working set from maintained repos, CLAUDE.md importing AGENTS.md via the @-syntax (a plain link does not load), .gitattributes eol=lf (the Windows-CI "Delete ␍" fix), prettier+eslint as a read-only CI gate, a weekly Trivy scan tuned to actionable findings, tag globs as a coarse filter backed by a mandatory in-job semver check, prerelease-safe latest, the verified release-please adoption path (token-cascade dispatch, the Actions-may-create-PRs setting, concurrency, issues-write), and label-categorized release notes.

sailingnaturali Skill

Use when an AI agent backed by MCP servers or tools misbehaves — claims data is missing/unavailable, returns 404s or "server unreachable", gives plausible-but-wrong answers, queries the wrong path, or leaks chain-of-thought / control tokens. Probe ground truth before trusting the agent's account of itself.

soapwong Bundle

生成羚羊投资研究全球市场周报。触发词：weekly prompt mode。收集calendar图、新题材、周会笔记，启动Tavily搜索，最终生成并保存周报Markdown。

janeeyre3007 Bundle

Autonomous optimizer for OpenCode skills inspired by Karpathy autoresearch. Use when the user wants skill quality scoring, trigger-boundary review, test-prompt design, optimization history, or iterative improvement of one or more SKILL.md files: "优化skill", "skill评分", "auto optimize skills", "skill质量检查", "skill review", "skill打分". Do not trigger when the user only wants to apply a domain skill to a normal task, edit opencode config/MCP/agents without changing a skill, or fix ordinary application code.

janeeyre3007 Bundle

Create, revise, evaluate, and package OpenCode skills. Use this whenever the user wants to make a new skill, clean up an existing skill, improve trigger accuracy, add evals, compare iterations, benchmark whether a skill is actually helping, or review overlap across the current skill set and tighten trigger boundaries. If the user says "turn this workflow into a skill", "improve this skill", "test this skill", "add evals", "optimize the description", or "check all current skills and micro-tune them", use this skill early. Do not trigger when the user only wants to apply a domain skill to a target, or when they are editing opencode config, agents, plugins, MCP servers, or permission rules without changing a skill definition.

janeeyre3007 Bundle

Firefox Reverse / frx-director / firefox-reverse-ai-mcp browser reverse workflow. Use when the user explicitly wants AI-driven browser reverse engineering through firefox-reverse-ai-mcp, frx-director-mcp, Firefox Reverse, or asks to choose between worker delegation and direct-drive browser engine tools. On activation, ask the user to choose the mode with the question tool unless the mode is already specified. Can coordinate with advanced reverse skills by full handoff or by loading their methodology while Firefox Reverse remains the execution engine. Do not trigger for WeChat miniapp debugging, pure AST deobfuscation, browser hook snippets only, Node/vm environment patching only, compact Python+iv8 scripts, or complete browser-free protocol collector implementation.

janeeyre3007 Bundle

生成浏览器 DevTools Console / Snippets 可直接粘贴执行的观察型 hook 脚本。追踪 cookie、XHR、fetch、header、storage、WebCrypto、canvas、Worker、Blob、DOM 注入等动态行为。适合用户已知观察目标、只缺可执行脚本时。默认最小侵入、页面层观察。当需要验证hook效果或定位脚本位置时，MCP使用策略：优先 cloakbrowser-reverse-mcp（启动脚本 E:\\BaiduNetdiskDownload\\JS逆向2622期\\JS逆向17 webpack打包方式 - 2-2026-6-3-顾安\\2026_6_3_webpack - 2\\2026_6_3_webpack - 2\\CloakBrowser\\cloakbrowser-reverse-mcp\\launch.bat，含 CDP/SourceMap/Profiler/WebSocket/Hook 能力）；只有需要 Camoufox 引擎级指纹/属性追踪证据时转 camoufox-js-reverse。不适合：不知道要hook什么，需要先定位函数位置（转 camoufox-js-reverse）、要脱离浏览器在 Node.js 跑（转 env-patch）、要 Python + iv8 脚本（转 iv8-web-reverse）。

janeeyre3007 Bundle

【优先级：微信小程序任务必选】微信小程序 / WeChat Mini Program / WMPF 逆向调试。强触发关键词：微信小程序、PC微信小程序、WMPF、WeChatAppEx、AppService、WebView、127.0.0.1:62000、WMPFDebugger、小程序调试。使用 miniapp-reverse-mcp 通过 CDP 调试端口分析小程序网络请求、源码、断点、调用栈和加密参数入口。不要被 camoufox-js-reverse 抢占。不要用于：普通 Web API 签名（转 camoufox-js-reverse 或 web-protocol-recovery）、浏览器 hook 脚本、Node.js 补环境、iv8 请求脚本、AST 解混淆或瑞数骨架任务。

olaservo Skill

Complete documentation for the Skilljack MCP server - tools, prompts, resources, configuration, and architecture reference.

dhicoc Bundle

五运六气（运气学）AI Agent 技能包。帮助人类准确理解《黄帝内经》运气学思想体系（天人合一、气化、中和、时间节律）。 提供干支推算、思想解读、概念解释、导出复习材料、自进化优化。 触发词：五运六气、运气、大运、主运客运、主气客气、司天在泉、客主加临、 太过不及、天符岁会、运气病机、运气治法、七篇大论。 面向：学生、医师（理论参考）、研究者。

dhicoc Bundle

刘完素（约1110—1200，金代）的思维框架与表达方式。河间学派创始人，寒凉派代表。 核心心智模型：六气皆从火化、亢则害承乃制（兼化是虚象不可误治）、审机求属。 决策启发式：见火热证先从火化推、兼化假象不误治本气、吐酸为热非寒、不可峻用辛温大热。 用途：作为运气学思维顾问，用刘完素的视角分析病机、审视治法、提供反馈。 触发词：「刘完素视角」「河间怎么看」「寒凉派」「六气皆从火化」「刘完素会怎么看」「切换到刘完素」「liu wansu perspective」。 即使用户只说「用河间的角度想想」「如果刘完素会怎么治」也应触发。 素材来源：公版《素问玄机原病式》《素问病机气宜保命集》全文（已蒸馏入 rag-knowledge-base/）。

dhicoc Bundle

张介宾（1563—1640，明代，字景岳）的思维框架与表达方式。温补派代表。 核心心智模型：阳气为本（命门真火）、五行互藏、生中有克克中有用、亢则害承乃制（造化不可无制）。 决策启发式：治病温养阳气为先、五行生克辩证看（生非纯生克非纯害）、客主但有胜而无复、天符岁会轻重分级。 用途：作为运气学思维顾问，用张介宾的视角分析病机、审视治法、提供反馈。 触发词：「张介宾视角」「张景岳怎么看」「温补派」「阳气为本」「张介宾会怎么看」「切换到张介宾」「zhang jiebin perspective」。 即使用户只说「用景岳的角度想想」「如果张介宾会怎么治」也应触发。 素材来源：公版《类经图翼》《类经》运气类全文（已蒸馏入 rag-knowledge-base/）。

dhicoc Bundle

从明·张介宾（张景岳）《类经图翼》卷一「运气上」、卷二「运气下」蒸馏出的运气学「图翼·象数基础」层框架——太极—阴阳—五行生成数—气数—五运（五天五运/五音建运太少相生/主客运）—六气（正化对化/主客气/司天在泉）—天符岁会—南北政脉不应，逐字照明本底本。用于「象数基础研读」与「图翼推算细节补足」，非临床指南。

dhicoc Bundle

从《运气证治歌诀》（清·王旭高/王泰林，辑自陈无择《三因方》司天运气诸方）蒸馏出的运气证治框架——五运十方、六气六方、运气推算歌诀、六淫治例。用于"按年干支回查运气方框架"与"运气治法研读"，非临床处方指南。

dhicoc Bundle

从宋·刘温舒《素问入式运气论奥》（元符己卯 / 1099，已入《正统道藏》太玄部，公有领域）蒸馏出的运气学「专论·机制纵深层」框架——卷上象数纵深（纳音／六化／五行本义／日刻／标本）+ 卷下机制纵深（胜复·五郁·六病·六脉·治法·五行胜复论＝亢害承制机制化），逐字照明本底本。用于「专论机制回溯」与「亢害承制／胜复／五郁／六病桥梁」补足，非临床指南。

dhicoc Bundle

从宋·陈言（陈无择）《三因极一病证方论》卷之五「五运时气民病证治」「六气时行民病证治」蒸馏出的运气用方「方源」层框架——十六方（五运十方 + 六气六岁方）的药物组成、炮制分量、主治、煎服法与随气加减，逐字照宋本底本。用于「按年干支回查运气用方」与「运气方源研读」，非临床处方指南。

dhicoc Bundle

从《医宗金鉴》卷三十五「编辑运气要诀」（清·吴谦等编纂）蒸馏出的运气推算全框架——太极阴阳、五行生克、五运六气、脏腑经络映射、天符岁会、南北政、运气为病。用于"按年干支回查运气格局与病机框架"与"运气推算研读"，非临床处方指南。

yeachan-heo Skill

Consensus planning entrypoint that auto-gates vague ultragoal requests before execution