Skills de Claude Code · página 114

>

>

>

>

Autonomy contract and routing index for autonomous project controllers. Loaded every tick; carries the rules that must survive cron-prompt rewrites.

Delegate coding/automation tasks to sandboxed.sh missions via the mcp_sandboxed_assistant_* MCP. Each mission runs in an isolated container (workspace) with a chosen agent profile and a self-contained prompt. Use this skill whenever the user wants to 'launch a mission', 'sandboxed', 'spawn a worker', or delegate a multi-step coding/research task that should run in a clean environment.

Check whether Claude Code, Claude Agent SDK, and Codex SDK have new versions on npm. Use when the user asks to check for SDK/CLI updates, e.g. "check sdk updates", "check claude code update", "有沒有更新", "檢查 SDK 版本".

Mandatory for every Nomi corrective change: user-reported bugs, regressions, CI-only failures, flaky tests, performance or security defects, review/audit findings, and compatibility failures in any production path. Classify one_off versus recurring before implementation. Recurring and high-risk repairs require a schema-v3 contract, shared enforcement boundary, structural prevention, dependency lifecycle decision, and changed regression evidence.

品牌宣传片 playbook。把产品文案/卖点做成一条「3 秒钩子 → 卖点 → 使用场景 → 行动号召」的短宣传片，分阶段执行（剧本审阅 → 拆镜 → 落画布 → 生成 → 排时间轴），每阶段暂停让用户审阅。

编辑创作区文档：读取当前内容，追加或替换文本，维护分镜描述格式。

非对抗性动作场景（跑酷/追逐/攀爬/特技/坠落）的动作编排知识——专业动作词汇、环境交互、动作节奏、FPV 追拍、身体力学约束、怎么把动作精确物理化地描述进视频提示词；Nomi 写动作戏 shot 时参考。

服化道——设计「人物设定图」与「场景环境图」的生图提示词，含顶部「风格前缀块」（摄影机/胶片/调色/画幅等烧进画面的统一风格）+ 生图 avoid（多指/文字水印/穿帮等）+ identity DNA（角色跨图一致的关键特征锁定）。Nomi 为角色/场景等视觉 anchor（`carrier: visual`）生成参考图、写它的生图 prompt 时参考。

镜头语言与摄影技法方法论——统一景别体系/构图规则/运镜的情绪语言/打光方案/景深控制/色温光源/镜头特性，以及这些怎么翻译成视频提示词该怎么写。Nomi 拆镜头或写视频 shot 的 prompt 时参考。

治 AI 视频多镜头「同角色换脸 / 同道具换形 / 同场景换景」三大顽疾的方法论——五维一致性检查、参考图锚定、场内状态表、段间承接。Nomi 拆镜头与跨镜生成时参考：把每个跨镜元素落成画布 anchor 节点（角色/场景/道具/风格），镜头用 anchorIds 引用、系统连参考边，段间用首尾帧承接。

中国古装题材形制知识——服饰造型(朝代/身份/发冠/妆容/色彩等级)、建筑形制(屋顶等级/开间/斗拱/院落/衙门公堂)、礼仪调度、穿帮自检，以及古装参考图/视频提示词要点(古匾汉字放开、现代穿帮 avoid)；Nomi 做古装短片/微短剧时参考。

参考图（角色/场景设定图）的 Go/No-Go 审图清单——逐项检查 + 判定规则 + 反馈模板，帮用户快速识别设定图是否合格、不合格时怎么反馈调整重抽；Nomi 生成参考图后审核时用。

非武打文戏的表演指导——微表情/肢体语言/视线设计/沉默表演/角色行为模式。AI 不会自己「演」，情绪必须在 prompt 里精确物理化成可拍的身体信号（眉/眼/嘴/手/肩/呼吸/停顿）。Nomi 写文戏 shot 的 prompt 时参考，让画面演得出「情绪」而不是只写情绪词。

把「镜头意图」翻译成视频提示词的手册——运镜翻译表（推/拉/摇/跟/环绕…→提示词措辞）、焦点翻译、复合镜头(多 clip)模式、污染词替换铁律、把抽象动作/情绪换成物理动作。Nomi 写视频 shot 的 `prompt` 字段时参考。

声音设计知识——环境音/音效/配乐情绪/沉默设计/听觉记号（leitmotif）/声画对位，仅供叙事规划与后期剪辑参考；⚠️视频生成提示词不含音频，这些内容不写进 shot 的 prompt，是给 Nomi 时间轴后期配音/配乐用的。

多角色调度方法论——3 人以上同框的站位设计/视线网络/注意力引导/前后景分层/AI 生成中的多角色一致性控制；Nomi 写多人 shot 的 prompt（谁站哪/朝向/视线/谁是焦点）时参考，走应用内 Seedance/可灵生成。

导演风格融合招式·大友启史 × 埃德加·赖特（时尚动作怪盗 DNA）——按情节类型（飙车/打斗/跑酷/盗窃潜入/文戏温情/时尚换装登场）分别配运镜 + 剪辑节奏 + 声画卡点；Nomi 做高燃动作 / 时尚怪盗戏、想要这种质感时套用。运镜写进镜头 prompt，卡点节奏当时间轴剪辑参考，经应用内 Seedance / 可灵生成。

创意转场方法论。AI 视频逐条生成、帧天然对不齐，硬切会露缝——用遮挡/甩镜/入暗/匹配剪辑/物体擦过镜头等创意转场把相邻两条缝合成一气呵成，同时把转场当叙事工具用（场景切换/时间跳转/角色替换/情绪转折）。Nomi 里设计跨镜衔接、给相邻 shot 写首尾帧提示词时参考。

短剧 playbook。把故事梗概/小说片段做成一条有人物、有转折的剧情短片，分阶段执行（剧本审阅 → 立圣经+拆镜 → 落画布冻结定妆 → 参考图生视频 → 排时间轴），每阶段暂停让用户审阅。跨镜不换脸靠「先冻结角色卡、再从它生成」。

Connect HTTP models or a native ComfyUI workflow through Nomi's verified certification path.

Nomi 桌面端发版素材包 playbook。先核版本事实和公开样片，再用故事串起核心更新，在 Nomi 中规划/生成/排片，最后诚实交接双语成片、更新图文、社交文案与本地 QA。

技能转写师。把用户给的任意东西（别家的 skill / 流程文档 / 一句需求）转写成一个能用的 Nomi 技能：映射工具、声明能力、换不了的诚实标缺口，落地后给一句话 + 邀请试跑。

创作区 AI 助手。支持普通写作、剧本创作、提示词生成、分镜描述。通过 documentAction 协议返回修改建议，用户确认后才写入。

定妆/定景规划师。从一段剧本里识别主要角色与关键场景，为每个建一张卡节点，并注入「身份板/场景板」提示词（剧本驱动、逐字标注、身份锁定），写入生成画布。

生成区 AI 助手。根据用户描述规划画布节点（图片/视频），只创建带提示词的节点，不执行生成。

故事板规划师。把用户给的故事规划成一份「分镜方案」结构化对象（跨镜头要一致的角色/场景/道具/风格 + 每个镜头的时长/参考/提示词），交给用户在创作区审阅、修改，确认后才落画布。

改编顾问——原著分析 / 改编策略 / 戏剧点提炼 / 结构映射 / 人物迁移。Nomi 创作区把小说/原著改编成剧本时参考。

行为心理学编剧工具——把认知失调 / 依恋理论 / 防御机制 / 微表情等心理学工具转化为角色行为设计与观众心理操控。Nomi 创作区设计角色行为/心理转折时参考。

对白专家（David Mamet 方法论）——台词创作 / 潜台词设计 / 角色语言个性化 / 对白节拍分析 / 沉默设计。Nomi 创作区写剧本/对白时参考。

即兴生台词——让角色在设定好的世界里即兴对戏，长出编剧预设不出的真实口语与情节；再把即兴原料按「剪水→留金→上结构→转场景」消化成可拍剧本。Nomi 创作区想要自然口语/真实关系张力时用。

创意孵化——在有故事骨架之前，基于 Truby 有机结构 + 宋方金人物先于情节 + Kasdan 核心信念检验，通过多轮追问帮用户从碎片里长出故事的根系（主题/人物内核/核心信念/世界）。Nomi 创作区立项初期、还没想清楚要写什么时用。

小说消化——把原著文本按 Truby/Mamet/McKee/Gilroy 方法论拆解，输出结构分析/对白范例/场景范例/叙事技法/本剧对标映射。Nomi 创作区改编或从原著取养分时参考，分层读取不必一次读完全本。

剧本审查·导演视角——用视觉叙事/节奏把控/镜头语言/空间调度/情绪曲线等导演维度审一份剧本，找出「读着顺但拍不出/立不住」的问题。Nomi 创作区写完剧本回头检查时用。

专业编剧（Lawrence Kasdan《星战》《肖申克》× Tony Gilroy《谍影重重》方法论）——故事结构/场景写作/对白/质量审查，含格式变体（长片/剧集/短片/超短）与严肃叙事风格。Nomi 创作区写剧本时参考。

宋方金编剧方法论（中国编剧《手机》《给青年编剧的信》）——人物先于情节 / 生活质感 / 关系即戏剧 / 对白零废话。Nomi 创作区写剧本时参考，尤其中式现实题材。

结构师（John Truby 方法论）——故事有机结构设计 / 对手网络构建 / 场景编织 / 节奏控制 / 钩子编排。Nomi 创作区搭故事骨架时参考。

保罗·施拉德（Paul Schrader，《出租车司机》编剧、《美国舞男》《第一归正会》导演）的叙事风格——超验电影式的极简、克制与蓄压，写孤独主角在平庸日常里被侵蚀、由偶遇而非选择推向转折，意义在"一切照旧"中悄然位移。适用于都市孤独、边缘人物、慢节奏内心戏、生活流的现实题材。Nomi 创作区想要这种冷峻克制、不解释情绪、靠并置而非因果说话的剧本质感时套用。

When writing code against an external library, framework, SDK, or API, look up its CURRENT documentation with the context7 MCP tools instead of relying on memory. Use whenever you import/call a third-party package or hit an unfamiliar or fast-moving API.

Before telling the user a coding task is done, actually verify it — build, run the relevant tests/linter, and confirm the change does what was asked with no regressions. Use whenever you are about to claim a change is finished, fixed, or working.

>

>

>

>

>

>

Audit and improve project-rules files (AGENTS.md, CLAUDE.md, .agents/instructions, local overrides) so the agent keeps accurate project context. Use when the user asks to check, audit, review, update, improve, or fix their AGENTS.md or CLAUDE.md, mentions "project rules maintenance" or "agent context optimization", or when the codebase has changed enough that the rules file may be stale. Scans the repository for every rules file, grades each against a quality rubric, outputs a quality report, and applies targeted edits only after user approval.

Capture learnings from the current session into the project-rules file (AGENTS.md, CLAUDE.md, or local override) so future sessions benefit. Use when the user says "revise the rules", "update AGENTS.md / CLAUDE.md with what we just learned", "save this to project memory", "remember this for next time", or at the end of a productive session when valuable context has emerged that is not yet documented. This complements agents-md-improver — improver audits, while this one captures.

Design a feature architecture by analyzing existing codebase patterns and conventions, then provide a comprehensive implementation blueprint with specific files to create or modify, component designs, data flows, and a build sequence. Use this skill when the user asks for an architecture design, an implementation plan for a non-trivial feature, or when dispatched as a sub-task during feature-dev architecture phase.

Deeply analyze an existing codebase feature by tracing execution paths, mapping architecture layers, understanding patterns and abstractions, and documenting dependencies. Use this skill when you need to understand how a feature works before modifying or extending it, when dispatched as a sub-task during feature-dev exploration, or when the user asks "how does X work in this codebase".

Review a pull request or a set of code changes for bugs, logic errors, and project-convention violations using a confidence-filtered, multi-agent process. Use this skill when the user asks to review a PR, audit pending changes, or inspect a diff for problems before merging.

Review code for bugs, logic errors, security vulnerabilities, code quality issues, and adherence to project conventions, using confidence-based filtering to report only high-priority issues that truly matter. Use this skill when reviewing a small set of changes locally (such as unstaged diff), when dispatched as a sub-task during feature-dev quality review, or when the user wants a critique of a specific file or function.

Guide a feature implementation through a structured seven-phase workflow with deep codebase understanding, clarifying questions, parallel architecture design, and quality review. Use this skill when the user asks to build a new feature, add functionality, or wants a methodical approach to implementation rather than diving straight to code.

Create distinctive, production-grade frontend interfaces with high design quality and accessible markup. Use this skill when the user asks to build or beautify web components, pages, applications, landing pages, dashboards, artifacts, or React/HTML/CSS UI. Generates creative, polished code that avoids generic AI aesthetics, then self-checks it against an objective accessibility and quality rubric.

Guide the creation of high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when the user wants to build an MCP server to integrate an external API or service, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).

Perform a focused security review of pending git changes to identify high-confidence security vulnerabilities with real exploitation potential. Use this skill when the user asks for a security review, security audit, vulnerability scan, or wants to check pending changes on a branch for security issues before merging. This is NOT a general code review.

Create new skills (SKILL.md files), modify and improve existing skills, and design skill descriptions for accurate triggering. Use when the user wants to create a new skill from scratch, edit an existing skill, optimize a skill's description, or convert a workflow they just demonstrated into a reusable skill.

Audit GitHub Actions that run AI agents for prompt injection, unsafe interpolation, sandbox gaps, and permissive actor rules. Use for agentic CI workflows, not general application code review.

Operational rubric that turns "don't make AI slop" into observable properties, severity levels, evidence requirements, and repair actions for interface design. Use as the reference rubric when building or reviewing marketing sites, product interfaces, dashboards, portfolios, or e-commerce pages, especially alongside frontend-design.

Agents should invoke this skill for code reviews, linting/formatting setup, maintainability checks, complexity concerns, warning cleanup, coding standards, or quality gates in Rust, TypeScript, Python, shell, and mixed repos.

Run CodeQL database creation and security queries, add data-extension models, or process CodeQL SARIF. Use when CodeQL is explicitly requested; use security-review for a broader manual security review.

Agents should invoke this skill when choosing patterns, designing traits/interfaces/components, deciding abstraction boundaries, evaluating dependency injection/callbacks, or comparing implementation approaches in Rust, TypeScript/React, or Django/Python.

Performs security-focused differential review of code changes (PRs, commits, diffs). Adapts analysis depth to codebase size, uses git history for context, calculates blast radius, checks test coverage, and generates comprehensive markdown reports. Automatically detects and prevents security regressions.

Systematically verifies suspected security bugs to eliminate false positives. Produces TRUE POSITIVE or FALSE POSITIVE verdicts with documented evidence for each bug.

Use the `hf` CLI for Hub authentication, downloads, uploads, repositories, cache, jobs, buckets, webhooks, and endpoint administration. Use for CLI operations, not model recommendations, paper analysis, training design, or building a Space.

Discover the active AWS profile, region, account, and caller identity before SageMaker or AWS work. Use when local AWS context is needed or unspecified; do not use for cloud-agnostic model planning.

Create an isolated Python environment with a compatible Python version and boto3 for SageMaker or AWS automation. Use before executing AWS Python code; do not use for general Python environment setup.

Plan and coordinate a model deployment to Amazon SageMaker, including serving stack and real-time versus async inference. Use as the entry point for SageMaker hosting requests, before image, IAM, and endpoint implementation skills.

Verify or select a SageMaker execution role before creating models, endpoints, or training jobs. Use when a role ARN is missing or IAM access fails; inspect existing roles before proposing role creation.

Implement a production SageMaker endpoint with autoscaling, CloudWatch alarms, and tags. Use after the serving image and IAM role are known; use the deployment planner first when architecture is undecided.

Select and verify the current region-specific serving container URI for a SageMaker model deployment. Use after the deployment pathway is chosen and before endpoint code; never infer an image URI from memory.

Hugging Face CLI to estimate the required memory to load Safetensors or GGUF model weights for inference from the Hugging Face Hub.

Find and compare recommended Hugging Face models for a task using benchmarks, model size, and device constraints. Use for model selection questions; use huggingface-local-models for GGUF setup and hf-cli for Hub operations.

Run evaluations for Hugging Face Hub models using inspect-ai and lighteval on local hardware. Use for backend selection, local GPU evals, and choosing between vLLM / Transformers / accelerate. Not for HF Jobs orchestration, model-card PRs, .eval_results publication, or community-evals automation.

Use this skill for Hugging Face Dataset Viewer API workflows that fetch subset/split metadata, paginate rows, search text, apply filters, download parquet URLs, and read size or statistics.

Build Gradio web UIs and demos in Python. Use when creating or editing Gradio apps, components, event listeners, layouts, or chatbots.

Train or fine-tune language models with TRL or Unsloth on Hugging Face Jobs, including SFT, DPO, GRPO, reward models, and GGUF conversion. Use for cloud LLM training; use huggingface-vision-trainer for vision tasks.

Use to select models to run locally with llama.cpp and GGUF on CPU, Mac Metal, CUDA, or ROCm. Covers finding GGUFs, quant selection, running servers, exact GGUF file lookup, conversion, and OpenAI-compatible local serving.

Build and publish a Gradio Hugging Face Space for a user-provided image or video LoRA, including pipeline selection and model-card settings. Use for LoRA demos; use huggingface-spaces for general Space work.

Publish and manage research papers on Hugging Face Hub. Supports creating paper pages, linking papers to models/datasets, claiming authorship, and generating professional markdown-based research articles.

Look up and read Hugging Face paper pages in markdown, and use the papers API for structured metadata such as authors, linked models/datasets/spaces, Github repo and project page. Use when the user shares a Hugging Face paper page URL, an arXiv URL or ID, or asks to summarize, explain, or analyze an AI research paper.

Build, deploy, debug, or maintain a Hugging Face Space using Gradio, Docker, or Static SDKs. Use for general Space hosting and configuration; use huggingface-zerogpu for ZeroGPU runtime constraints and lora-space-builder for LoRA demos.

Use this skill when the user wants to build tool/scripts or achieve a task where using data from the Hugging Face API would help. This is especially useful when chaining or combining API calls or the task will be repeated/automated. This Skill creates a reusable script to fetch, enrich or process data.

Track and visualize ML training experiments with Trackio. Use when logging metrics during training (Python API), firing alerts for training diagnostics, or retrieving/analyzing logged metrics (CLI). Supports real-time dashboard visualization, alerts with webhooks, HF Space syncing, and JSON output for automation.

Train object-detection, image-classification, or SAM segmentation models on Hugging Face Jobs. Use for vision fine-tuning and evaluation; use huggingface-llm-trainer for language models.

Implement or debug Hugging Face Spaces ZeroGPU code using `@spaces.GPU`, including isolation, state, duration, sizing, and CUDA-wheel constraints. Use for ZeroGPU-specific behavior; use huggingface-spaces for general hosting.

Detects fail-open insecure defaults (hardcoded secrets, weak auth, permissive security) that allow apps to run insecurely in production. Use when auditing security, reviewing config management, or analyzing environment variable handling.

Agents should invoke this skill for academic or technical papers, arXiv/PubMed/IEEE/ACM links, PDFs, methodology review, limitations, practical implications, or extracting findings for engineering decisions.

Parses and processes SARIF files from static analysis tools like CodeQL, Semgrep, or other scanners. Triggers on \"parse sarif\", \"read scan results\", \"aggregate findings\", \"deduplicate alerts\", or \"process sarif output\". Handles filtering, deduplication, format conversion, and CI/CD integration of SARIF data. Does NOT run scans — use the Semgrep or CodeQL skills for that.

Create a repository-grounded AppSec threat model covering assets, trust boundaries, attackers, abuse paths, and mitigations. Use only for explicit threat-modeling requests, not general architecture or code review.

Creates custom Semgrep rules for detecting security vulnerabilities, bug patterns, and code patterns. Use when writing Semgrep rules or building custom static analysis detections.

Creates language variants of existing Semgrep rules. Use when porting a Semgrep rule to specified target languages. Takes an existing rule and target languages as input, produces independent rule+test directories for each language.

Run Semgrep static analysis across a codebase, optionally using Semgrep Pro for cross-file taint analysis. Use when Semgrep or a static-analysis scan is requested; use security-review for a manual audit.

Identify error-prone APIs, dangerous configuration, and footgun designs that enable security mistakes. Use for API ergonomics, misuse resistance, secure defaults, or pit-of-success reviews; use code-review for general defects.

Identifies dependencies at heightened risk of exploitation or takeover. Use when assessing supply chain attack surface, evaluating dependency health, or scoping security engagements.