Skills 目录 · 第 48 页

Translate owned plainspoken text artifacts one way from United States English (en-US) to France French (fr-FR), preserving source voice, meaning, protected tokens, structure, and the pair-specific dictionary. Use only after any specialist register has separately been mediated into plain en-US. Do not use for another pair, a register conversion, reverse translation, generic copyediting, or machine-only publication.

Review one or more project inbox items and assign a recorded lifecycle disposition. Use for inbox sweeps, duplicate resolution, grouping related captures, activating an effort, identifying a graduation candidate, archiving, or requesting owner input. Do not publish or move work outside the project.

Draft an X reply, called a Twitter comment in this family, to a supplied post and conversation. Use when the user wants to reply on X, answer a tweet, join an existing X thread, request feedback, or report public progress. It keeps the parent context and reply visibility in view, then returns a draft only. Do not use for a new standalone post or publication.

Draft an X post using the portable name Twitter. Use when the user asks for a tweet, X post, concise Twitter update, quote text, standalone build log, or a thread opener. It writes one evidence-locked post for a named account and returns a draft only. Do not use for replies, posting automation, or assuming X Premium features.

Generate and refresh Mermaid visual sitemaps from search indexes. Use when a site's universe map drifts after crawling, indexing, or adding pages, or when multiple sibling sites need consistent navigation diagrams. Includes a Python generator, templates, good and bad examples, and Mermaid tool guidance. Does not infer project completion from page existence or deploy sites.

Fetch or integrate scripture passages from Judaism, Christianity, or Islam through the ARE free anonymous providers. Use for verse lookup, API endpoint and reference-format questions, translation IDs, normalization, attribution, error handling, fallback routing, or denomination-aware scripture UI work. Covers Sefaria, bible-api.com, Quran.com with AlQuran.cloud fallback, and LDS Standard Works routing. Never invent text or silently change translations.

Generate, validate, normalise, and explain Mermaid-native bpmn-beta diagrams from a Process Narrative Specification. Use this skill when the user wants to convert a PNS or process notes into a visual diagram; when they ask for BPMN, a process flow, a swimlane diagram, or a workflow diagram; when they want to validate or repair existing bpmn-beta code; when they mention pools, lanes, gateways, tasks, events, or BPMN 2.0; when they want a Mermaid-native process model they can commit to a repo; when the PNS has three or more gateway decision points and a DMN decision table is warranted. Do not use it for a non-BPMN diagram type (flowchart, sequence, ER) — use standard Mermaid syntax for those instead. Produces bpmn-beta.mmd and, optionally, a rendered process-model.svg.

OverKill Hill P³ Vite-to-GitHub-Pages deployment runbook. Use when deploying or troubleshooting this React/Vite app, its Actions workflow, /abrahamic-reference-engine/ production base, BrowserRouter basename fallback, dist artifacts, build validation, or Pages environment variables. Also activate for subpath asset failures or SPA refresh diagnosis. Do not use for gh-pages branches, backend hosting, or unrelated UI work.

Create or renew an evidence-grounded voice, tone, and glossary knowledge file from an authorized writing corpus. Use for a baseline, quarterly or annual update, or correction across audiences. It does not write a social post, certify identity, infer private traits, or publish a profile.

Draft a YouTube comment or reply from supplied video and conversation context. Use when the user wants to respond to a viewer, answer feedback, add a correction, or join a discussion below a video or channel post. It returns draft text only and does not reply, heart, pin, hide, or moderate comments.

Draft a YouTube channel post or participant Community post from supplied destination context. Use for text, image, video, poll, or quiz posts that need a platform-native prompt and audience boundary. It prepares a draft only and does not test eligibility, schedule, or publish.

Draft a YouTube video publishing brief and metadata bundle from supplied video facts. Use when the user needs a title, description, chapters, CTA, or disclosure and setting handoff for a YouTube video or Short. It prepares text only and does not upload, edit, schedule, or publish a video.

Build Solana trading and intelligence applications combining OKX DEX aggregation with Helius infrastructure. Integration-only layer — describes when and how to compose OKX tools with Helius tools for swaps, token discovery, smart money signals, meme token analysis, and portfolio intelligence.

Build Solana trading and intelligence applications combining OKX DEX aggregation with Helius infrastructure. Integration-only layer — describes when and how to compose OKX tools with Helius tools for swaps, token discovery, smart money signals, meme token analysis, and portfolio intelligence.

显微成像数据管理平台。通过 Python 访问图像、获取数据集、分析像素、管理 ROI/标注、批量处理，适用于高内涵筛选和显微成像工作流。

以无头方式运行耗时较长的 OMP（`omp -p`）任务，使其在派发它的会话结束后仍能继续运行——使用分离的 tmux、NDJSON 事件日志、轮询完成状态、会话 ID 恢复。适用于委派需要运行数分钟以上的仓库任务、编排用的聊天/会话可能重启，或恢复在回合中途被终止的运行。不适用于两分钟即可完成的命令（直接内联运行），也不适用于需要实时操控的工作（那属于 TUI 会话）。

Use for /cs-onboard, or when the virtual C-suite lacks company context. Interviews the founder and writes ~/.claude/company-context.md.

当注册后的激活、实现价值所需时间或首次使用体验需要改进时，在重新设计之前使用。通过 activation_funnel_analyzer.py 找出激活流失点。

当用户希望优化注册后的引导流程、用户激活、首次使用体验或价值实现时间时使用。也适用于用户提到“引导流程”“激活率”“用户激活”“首次使用体验”“空状态”“引导清单”“顿悟时刻”或“新用户体验”时。关于注册/登记优化，请参阅 signup-flow-cro。关于持续的邮件序列，请参阅 email-sequence。

Create or update standardized object-oriented component documentation using a shared template plus mode-specific guidance for new and existing docs.

通过 open-banking.io 的 PSD2 API 读取欧盟/英国银行的账户余额和交易。适用于用户希望查看余额、列出近期交易、对支出分类，或在欧洲银行账户之间核对付款——无需 eIDAS 证书或 AISP 许可证。

Open the pull request for the branch you are working on in the browser via `gh pr view --web`. Trigger when the user says: "PRを開いて", "PRをブラウザで", "PRを見せて", "open pr", "open the PR", "show me the PR", or similar phrases requesting to open the working pull request in the browser.

Use when the user asks how to build with OpenAI products or APIs and needs up-to-date official documentation with citations (for example: Codex, Responses API, Chat Completions, Apps SDK, Agents SDK, Realtime, model capabilities or limits); prioritize OpenAI docs MCP tools and restrict any fallback browsing to official OpenAI domains.

使用OpenAlex数据库查询和分析学术文献。当搜索学术论文、分析研究趋势、查找作者或机构的著作、追踪引用、发现开放获取出版物，或在240M+学术著作中进行文献计量分析时，应使用此技能。适用于文献检索、研究成果分析、引用分析和学术数据库查询。

Generate a complete, production-ready application from an OpenAPI specification

将 OpenAPI 3.0 JSON/YAML 转换为 TypeScript 接口和类型守卫。当用户要求从 OpenAPI 生成类型、将 schema 转换为 TS、创建 API 接口，或根据 API 规范生成 TypeScript 类型时，应使用此技能。

为 OpenClaw AI Agent 锻造完整的龙虾灵魂方案。根据用户偏好或随机抽卡， 输出身份定位、灵魂描述(SOUL.md)、角色化底线规则、名字和头像生图提示词。 如当前环境提供已审核的生图 skill，可自动生成统一风格头像图片。 当用户需要创建、设计或定制 OpenClaw 龙虾灵魂时使用。 不适用于：微调已有 SOUL.md、非 OpenClaw 平台的角色设计、纯工具型无性格 Agent。 触发词：龙虾灵魂、虾魂、OpenClaw 灵魂、养虾灵魂、龙虾角色、龙虾定位、 龙虾剧本杀角色、龙虾游戏角色、龙虾 NPC、龙虾性格、龙虾背景故事、 lobster soul、lobster character、抽卡、随机龙虾、龙虾 SOUL、gacha。

为 OpenClaw AI Agent 锻造完整的龙虾灵魂方案。根据用户偏好或随机抽卡， 输出身份定位、灵魂描述(SOUL.md)、角色化底线规则、名字和头像生图提示词。 如当前环境提供已审核的生图 skill，可自动生成统一风格头像图片。 当用户需要创建、设计或定制 OpenClaw 龙虾灵魂时使用。 不适用于：微调已有 SOUL.md、非 OpenClaw 平台的角色设计、纯工具型无性格 Agent。 触发词：龙虾灵魂、虾魂、OpenClaw 灵魂、养虾灵魂、龙虾角色、龙虾定位、 龙虾剧本杀角色、龙虾游戏角色、龙虾 NPC、龙虾性格、龙虾背景故事、 lobster soul、lobster character、抽卡、随机龙虾、龙虾 SOUL、gacha。 Use when creating, designing, or customizing an OpenClaw lobster persona — identity, SOUL.md, name, or avatar prompt.

オープンソースパイプライン: プライベートプロジェクトをフォーク、サニタイズし、安全な公開リリースのためにパッケージ化する。3つのエージェント（フォーカー、サニタイザー、パッケージャー）を連鎖させる。トリガー: '/opensource'、'open source this'、'make this public'、'prepare for open source'。

开源流水线：对私有项目进行分叉、清理并打包，以便安全公开发布。串联 3 个智能体（分叉器、清理器、打包器）。触发词：'/opensource'、'open source this'、'make this public'、'prepare for open source'。当私有项目必须被分叉、移除机密信息并打包以便公开发布时使用。

开源流水线：fork、清理并打包私有项目以安全公开发布。串联3个代理（fork代理、清理代理、打包代理）。触发词：'/opensource'、'open source this'、'make this public'、'prepare for open source'。

为 /la:pr 流程在当前仓库初始化或修复 OpenSpec。确保 openspec CLI 可用，优先全局安装，npx 作为后备；使用 `openspec init --tools none` 创建 openspec/ 结构，仅建立语料库，不添加 /opsx:* 命令，并完成验证。当 openspec/ 缺失或无效时由 /la:pr 调用，也可直接运行。

查询 Open Targets Platform 的靶点与疾病关联、药物靶点发现、可成药性/安全性数据、遗传学/组学证据及已知药物，以识别治疗靶点。

用于 Flex/OT-2 机器人的实验室自动化平台。编写 Protocol API v2 协议、液体处理、硬件模块（加热振荡器、热循环仪）、实验室器皿管理，用于自动化移液工作流。

操作员审批契约，包含针对智能体起草的外发消息的内部归档通知、草稿哈希、按epoch键控的决策、持久投递声明与回执，以及起草前基线门禁。适用于智能体起草发给外部对手方的消息、且每条消息发出前必须由人工操作员批准、拒绝或引导的场景。

构建机会解决方案树（OST）以组织产品发现——将期望成果映射到机会、解决方案和实验。基于 Teresa Torres 的《Continuous Discovery Habits》。适用于组织发现工作、将机会映射到解决方案或决定下一步构建什么时。

在有关定时任务、修复或部署的说法送达操作者或队友前，将其转化为证据：找到真正记录该主张的界面或系统，区分手动复现与定时执行，并把外部审计者的结论视为一组待核查主张。用于验证 cron 结果、“已修复”声明、CI/审查机器人关卡，或环境批次状态报告。不用于审计书面报告内容，也不用于验收队友交付；这些需另行检查。

Activation-aware weight quantization for 4-bit LLM compression with 3x speedup and minimal accuracy loss. Use when deploying large models (7B-70B) on limited GPU memory, when you need faster inference than GPTQ with better accuracy preservation, or for instruction-tuned and multimodal models. MLSys 2024 Best Paper Award winner.

将大语言模型量化为 8 位或 4 位，实现 50-75% 的内存减少且精度损失极小。适用于 GPU 内存受限、需要容纳更大模型或希望加快推理时。支持 INT8、NF4、FP4 格式、QLoRA 训练和 8 位优化器。可与 HuggingFace Transformers 配合使用。

Optimizes transformer attention with Flash Attention for 2-4x speedup and 10-20x memory reduction. Use when training/running transformers with long sequences (>512 tokens), encountering GPU memory issues with attention, or need faster inference. Supports PyTorch native SDPA, flash-attn library, H100 FP8, and sliding window attention.

GGUF格式与llama.cpp量化，用于高效的CPU/GPU推理。在消费级硬件、Apple Silicon上部署模型，或需要无需GPU的2-8位灵活量化时使用。

Post-training 4-bit quantization for LLMs with minimal accuracy loss. Use for deploying large models (70B, 405B) on consumer GPUs, when you need 4× memory reduction with <2% perplexity degradation, or for faster inference (3-4× speedup) vs FP16. Integrates with transformers and PEFT for QLoRA fine-tuning.

无需校准数据的 LLM 半二次量化。适用于在无需校准数据集的情况下将模型量化为 4/3/2 位精度、快速量化工作流，或使用 vLLM 或 HuggingFace Transformers 部署时。

端到端编排构建一个全新功能——调研、规划、TDD 实现、评审和受控提交——将每个阶段委派给匹配的 ECC 智能体。适用于添加尚不存在的能力时。

编排从设计或规范文档引导出可用的MVP——摄取SDD/PRD，规划薄垂直切片，搭建第一个端到端切片，然后驱动生成器-评估器构建循环，进行审查和门控feat提交。当设计或规范文档必须通过规划的垂直切片成为可运行的MVP时使用。

编排将现有可用的功能改为新的期望行为——将其测试更新到新规范，修改实现以匹配，审查，并进行门控提交。当行为未损坏但应当不同时使用。

将每个阶段委派给匹配的 ECC 智能体，编排缺陷修复：复现为失败的回归测试、修复至通过、审查，并经关卡检查后提交。现有行为损坏或错误时使用。

Shared orchestration engine behind the orch-* skill family — the gated Research-Plan-TDD-Review-Commit pipeline, size classifier, agent and command map, and two human gates (plan approval, commit confirmation) that orch-* operation skills delegate to. Use indirectly via orch-add-feature, orch-fix-defect, orch-change-feature, orch-refine-code, or orch-build-mvp; read directly only when adding an orch operation or tuning shared phases.

编排保持行为的重构——确认测试通过、在不改变行为的前提下重构、保持测试通过、审查，并受门控地提交。当结构应改进但行为不得改变时使用。

使用 Claude 或 Codex 智能体来编排委派的研究或实现工作。在 Claude Code 中默认使用 Claude，在其他所有执行框架中默认使用 Codex。遵循用户对智能体或模型的明确选择。无需单独的方案审批步骤即可规划并启动。