Claude Code Skills · page 52

催化剂与风险的反证式写法:每个强结论必须先找反证;催化剂按"兑现型 / 预期型 / 周期型"分类并要求可验证的数据时点;风险按技术路线断层、客户集中、产能过剩与价格战、周期顶、预期透支(假便宜 PEG)、一致预期下修、治理与流动性、数据源冲突分类;裁决点的标准写法(什么数据出来会改变判断 + 下一个公开数据时点);知识档案旧结论的反证处理。当任务涉及风险、反证、催化剂、裁决点、预期兑现、什么会推翻结论时加载;单纯取数、估值计算、财报拆分等不需要反证框架的任务不要加载。只产出框架、概率与裁决点,不给投资动作建议。

A 股个股研究六阶段 SOP(profile → financials → estimates → valuation → risk → report),Phase 0 范围 = 财务估值闭环。当任务是研究 / 分析 / 评估一只或多只**已指定代码**的 A 股个股时使用;规定每阶段取哪些数据、调哪些 calc 函数、必须落盘什么产物、过什么 Gate。不用于:从市场中筛选标的、泛行业讨论、概念解释、给投资动作建议。

A 股零鉴权取数手册。当需要真实的行情 / 市值 / 估值快照、季度报告期累计财务数据、机构一致预期 EPS、PE 历史序列、公告标题、日 K 线、交易日历时使用;只允许运行本 skill 登记的脚本取数(腾讯 / 新浪 / 同花顺 / baostock / 深交所 / 东财),禁止凭模型记忆给数,禁止自造爬虫。概念解释、观点讨论等不需要取数的话题不要加载。

财报拆解手册:报告期累计值 → 单季(quarterize)→ 最新单季 / TTM / TTM 同比 / 环比的口径地图,扣非与归母的取舍(一次性损益),季节性与报告期对齐(分子分母同期),三表交叉核对(利润表 / 资产负债表 / 现金流量表),比率(毛利率 / 费用率 / 负债率)一律经 calc ratio,"转向看最新期、规模看 TTM"的判读模板与质量检查清单。当任务涉及财报、季报、业绩、利润拆分、同比环比、毛利率、现金流、扣非时加载;只查行情 / 公告 / 产业链结构、或只讨论概念不涉及财务数字的任务不要加载。取数层不做任何算术,所有派生数字出自 calc;不给投资动作建议。

产业链下钻与不可替代性判定方法:以龙头为"需求入口"沿供应链逐层下钻(整机 / 龙头 → 部件 → 核心器件 → 材料 → 衬底与设备),用物理 / 材料约束(扩产周期、良率、认证周期、有无替代)当筛子找供给刚性的卡口;给每个标的贴不可替代性标签(tech_moat / capacity_moat / both / 待补)并列证据;含"卡口越硬越贵"与预期差四问的校准。当任务涉及产业链位置、上下游、护城河、不可替代性、供给瓶颈、竞争格局时加载;单纯取数、估值计算、财报拆分等不涉及产业链结构的任务不要加载。只产出框架与证据表,不给投资动作建议。

成长股估值口径手册(A 股为主,US/HK 通用):扣非×4 年化 PE、前瞻 PE、TTM PE 历史分位、PEG(扣非×4 PE ÷ 前瞻 CAGR)、前瞻 CAGR 与 TTM 同比交叉验证、一致预期分歧、四锚 PE 消化年数与"30 倍锚三铁律"、判读规则与常见错误。当任务涉及估值、PE、PEG、贵不贵、能不能消化、历史分位、一致预期时加载;只讨论概念、与估值无关的取数 / 行情 / 公告问题不要加载。所有数字一律经 calc/ 计算,本 skill 只管口径与判读,不给价格锚、不给投资动作建议。

Diagnose and configure MemSearch memory behavior. Use when the user asks about MemSearch configuration, plugin summarization, PROJECT.md/USER.md maintenance, memory directories, index health, provider routing, prompt files, or migration/compatibility questions.

Search and recall relevant memories from past sessions via memsearch. Use when the user's question could benefit from historical context, past decisions, debugging notes, previous conversations, or project knowledge -- especially questions like 'what did I decide about X', 'why did we do Y', or 'have I seen this before'. Also use when you see `[memsearch] Memory available` hints injected via SessionStart or UserPromptSubmit. Typical flow: search for 3-5 chunks, expand the most relevant, optionally deep-drill into original transcripts via the anchor format. Skip when the question is purely about current code state (use Read/Grep), ephemeral (today's task only), or the user has explicitly asked to ignore memory.

Turn workflows from your MemSearch memory into reusable skills. Use when the user asks to make/create/extract/distill a skill from what they just did or from past work, review skill candidates, install a distilled skill, or 'turn this into a skill'. Manages MemSearch procedural-memory candidates under .memsearch/skill-candidates/, not the host agent's own skills system.

Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.

Create, analyze, proofread, and modify Office documents (.docx, .xlsx, .pptx) using the officecli CLI tool. Use when the user wants to create, inspect, check formatting, find issues, add charts, or modify Office documents.

Create Super Agent Party (SAP) extensions. This skill should be used when users want to create, build, or scaffold a new extension for Super Agent Party - including static HTML extensions (pure frontend) and Node.js backend extensions. Triggers on requests like "create a new SAP extension", "build an extension for Super Agent Party", "scaffold a plugin", "make a chat UI extension", or when working with sap extension projects.

Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.

Migrate three legacy Markdown docs into Starlight-compatible migration candidates inside an isolated temp repo. Use when you need stage receipts, crash recovery, resume validation, or deterministic candidate packaging without touching real user docs.

Repair synthetic MDX frontmatter when title, description, or sidebar.order drift from the frontmatter contract. Use for files that follow the same observable patterns proven in the receipts; do not use on private transcripts, secret-bearing material, or arbitrary YAML.

Create or rewrite exactly one LearnPrompt Starlight MDX tutorial to the repository's golden-sample standard, with bottom-of-page auditable sources instead of SourceCard, a research pack, evidence ledger, reproducible showcase, independent review, quality score, and successful site build. Use for one LearnPrompt article, tutorial, chapter, or MDX page; in batch goals, invoke once per topic in an isolated task.

Save a shared web link as a clean Obsidian note. Use when the user gives a URL and asks to save, clip, or archive it into their Obsidian vault as a markdown note with title, source, and summary.

Distill a candidate skill from observable repair receipts, then prove it on holdout fixtures. Use when you have repeated accepted patches, corrections, and validator results, but do not have permission to use private transcripts or hidden chain-of-thought.

Prepare a dry-run rename plan for a batch of receipt PDFs. Use when the user wants repository-style receipt filenames normalized under the local naming policy without renaming files yet.

从一个 issue 和一段 diff 生成结构化发布摘要，固化“读 issue → 读 diff → 归类改动 → 评估风险 → 写验证步骤”的重复工作流。

Read a read-only release feed over HTTP, enforce an auditable request contract, and write redacted JSON and Markdown reports. Use when the task is to inspect releases from a REST endpoint without mutating upstream state.

Inspect a small npm CLI release candidate and produce a row-by-row release-readiness checklist report with evidence, pass rules, severity, N/A policy, and a final gate. Use when the task is to review a release candidate without publishing it.

Run any question, idea, or decision through a council of 5 AI advisors who independently analyze it, peer-review each other anonymously, and synthesize a final verdict. Based on Karpathy's LLM Council methodology. MANDATORY TRIGGERS: 'council this', 'run the council', 'war room this', 'pressure-test this', 'stress-test this', 'debate this'. STRONG TRIGGERS (use when combined with a real decision or tradeoff): 'should I X or Y', 'which option', 'what would you do', 'is this the right move', 'validate this', 'get multiple perspectives', 'I can't decide', 'I'm torn between'. Do NOT trigger on simple yes/no questions, factual lookups, or casual 'should I' without a meaningful tradeoff (e.g. 'should I use markdown' is not a council question). DO trigger when the user presents a genuine decision with stakes, multiple options, and context that suggests they want it pressure-tested from multiple angles.

A privacy-first, local-first search assistant and MCP server for your Zotero library, enabling AI agents to search and analyze your research papers securely.

Generate, edit-from-reference, or analyze images with AI via OpenRouter (gemini, geminipro, riverflow, flux2, seedream, gpt5, gpt5.4; Cloudflare AI Gateway BYOK). Also analyze a video (--analyze-video, read-only — no video generated) into a text description for video prompts. Use when the user asks to generate an image, create a PNG, make an icon, make it transparent, edit with a reference, design a logo/banner, describe/analyze/explain an image ("what's in this image"), or describe/analyze a video ("what happens in this video").

>

Consult official Claude Code documentation from code.claude.com using selective fetching. Use when working on hooks, skills, subagents, plugins, agent teams, MCP servers, permissions, settings, CI/CD (GitHub Actions, GitLab), IDE extensions (VS Code, JetBrains), desktop/web app features, scheduling, memory/CLAUDE.md, deployment (Bedrock, Vertex, Foundry), sandboxing, monitoring, or any Claude Code feature requiring official docs. Fetches only the specific docs needed per task.

Dual-AI code analysis pairing OpenAI Codex with Claude code-searcher — the lightest consult variant, two citation-verified perspectives. Use for a quick second opinion on a code question.

Dual-AI code analysis pairing z.ai GLM 5.2 with Claude code-searcher — a lightweight two-model second opinion. Use for a quick z.ai-backed check on a code question.

>

>

Cross-system agent token/cost audit (Claude Code, Codex, OpenClaw, Hermes, OpenCode): idle burns, model misrouting, config bloat, with dollar savings.

Pull a prior session's checkpoint on demand when the user is continuing prior work. Returns fenced, source-labeled, scrubbed recovery context. Do NOT call on a fresh, unrelated task.

Plan a token-efficient Claude Code or Codex setup, or get a quick health check. Coaching, not the full audit (use token-optimizer for that).

Open the Token Optimizer dashboard in your browser (context usage, quality, savings). Use to view the dashboard.

Audit a Claude Code or Codex setup for context-window waste, then fix it and measure the savings. Use when context feels tight.

Plan AI short films with story, shots, prompts, and continuity.

Use when acquiring or importing media into a OpenChatCut project asset library for video editing or creation, including local/attached videos, user-provided paths, public media URLs, web video/audio/image assets, upload fallback decisions, and deciding between import_media, download_media, or manual user action.

Use whenever the agent needs to add, create, hand-author, patch, or place Motion Graphic JSX assets in a OpenChatCut project. This is the direct-authoring path: use create_motion_graphic_from_code / edit_asset / edit_item, not motion-graphic-gen or submit_motion_graphic. Covers project/timeline intake, project visual language, editable properties, asset binding, inline JSX authoring, existing asset updates, timeline placement, and verification.

Create finished explainer videos from a topic, script, outline, voiceover, product logic, data, technical concept, course material, or reference assets. Use when the user wants narration, motion graphics, stock footage, generated visuals, or mixed visuals to explain an idea.

Use when a OpenChatCut video editing or creation workflow needs export, render, download, share, final delivery, subtitle-file export, render choice, local-only asset handling, or export fallback explanation.

|

Use when a OpenChatCut tool call fails or returns an unexpected shape.

Cut an imported livestream recording of any genre into evidence-backed, platform-ready clips by combining transcript, visual, audio, interaction, and domain-specific signals. Use for commerce, gaming, talk, interview, education, entertainment, sports, music, IRL, creative, news, or mixed livestream recordings.

Cut one long podcast, interview, course, livestream, or other source video into social-ready shorts, reels, highlights, or clip timelines from existing project media. Use when the user asks to cut a long video into Shorts, Reels, TikToks, Xiaohongshu posts, best moments, highlights, or multiple clips.

Add motion graphics at the right moments without blocking the story. Use when the user wants to enhance talking-head, lecture, tutorial, interview, podcast, or creator videos with motion graphics.

Turn multiple product shots, event footage, travel clips, gameplay moments, UGC/product footage, B-roll, or mixed media into social-ready reels, highlights, recaps, or montage-style short videos from existing project media.

|

|

新闻素材智能粗剪——把新闻素材粗剪为一条内容完整、逻辑清晰、节奏紧凑的新闻短视频。Use when the user asks to 粗剪新闻、新闻剪辑、把新闻素材剪成短视频、智能粗剪、news rough cut、编辑新闻视频, or provides news footage (发布会/采访/现场/监控素材) to be cut into a factual news short. Never add music, voiceover, or sound effects.

Connect an MCP-capable coding agent to OpenChatCut and edit local video projects. Use when the user asks to install, connect, or set up OpenChatCut; inspect or edit an OpenChatCut project; work with its timeline, transcript, captions, media, generation, motion graphics, audio, color, or export tools; or recover from an OpenChatCut MCP error.

Use for video editing or video creation work that should be editable in OpenChatCut, even when the user does not explicitly mention OpenChatCut. Covers local/attached video editing, captions/subtitles, transcription, trimming, talking-head cleanup, highlights, B-roll, overlays, generation, export, project/editor opening, importing, targeting, verifying, watching, and identifying the active OpenChatCut project/editor URL.

Turn a product into ad angles, hooks, scenes, and CTA.

|

|

Create or improve OpenChatCut custom skills. Use whenever the user asks to create a skill, capture a workflow as a skill, write a SKILL.md, add a reusable workflow, 创建技能, 把流程做成技能, 写一个技能, or improve an existing custom skill. Pushy: treat any workflow the user repeats as a skill candidate, even if they do not say "skill".

Break down each shot and turn the analysis into a storyboard reference. Use when the user wants shot-by-shot film analysis, director logic, cinematography breakdown, or a storyboard-style reference from a video.

|

Use when a video/audio task needs OpenChatCut transcription, captions, subtitles, subtitle styling, transcript search, transcript readiness checks, or enabling captions, including local or attached videos where the user asks to add captions/subtitles, transcribe, create bilingual subtitles, clean talking-head speech, remove filler words, or trim pauses.

Use when checking whether agent edits are reflected in the OpenChatCut project and editor.

|

Create platform-ready thumbnails from real video frames. Use when the user wants a thumbnail, cover image, YouTube cover, Shorts cover, Bilibili cover, Xiaohongshu cover, or other video poster image.

|

Use when the agent should ask the user for structured input with an in-chat form, including single-select, multi-select, text fields, style pickers, or voice audition choices.

Design and ship a multi-agent system on AgentField. Use when the user asks to build, scaffold, design, or run an agent, reasoner network, multi-agent backend, or 'an agent that does X' — whenever the work would otherwise be a single LLM call or a flat LangChain/CrewAI/AutoGen chain. The skill produces composite intelligence: a deep, dynamic, parallel reasoner graph with a working `docker compose up` smoke test. For an agent installed on this machine through `af` and visible in AgentField Desktop, use the `agentfield-personal` skill instead.

Build and install a personal AI agent on this machine's AgentField: real source in ~/agentfield-agents, packaged with agentfield-package.yaml, installed with `af install`, started with `af run`, registered on the local control plane, and visible in AgentField Desktop with a keys form and an auto-start toggle. Use when the user wants an agent that lives on their machine as a persistent capability — a pricing agent, a support agent, a research agent — rather than a deployable project. A standalone repository with Docker Compose is the `agentfield` skill; calling agents that already exist is the `agentfield-use` skill.

Whenever you have a discrete task to perform — one the user delegated, or one that arose inside your own work — check FIRST whether an installed AgentField agent covers it, and offload to it by default when one does. Coverage, not task size, is the test: even a small job goes to a covering agent. The check is cheap — one `af ls -e` or capability lookup answers it, and the roster can be cached for the session. These agents are your AgentField subharnesses: workers running on cheaper open models, in parallel, off your own context, every run recorded on the control plane and watchable at <server>/ui/runs/<run_id>. This skill cannot know what your fleet does — runtime discovery is the only source of truth (swe-planner, pr-af, and whatever else the user installed). Also the reactive path when the user asks to use, call, or run an agent, to list what agents or reasoners exist, or to check on an execution. Resolves the right control plane (local or desktop-configured cloud) first, fetches the reasoner's exact contract before dispatching, and calls entry-point reasoners only. Not for building new agents — that is the agentfield skill.

Add or update YAML frontmatter properties to enhance Obsidian note organization. Use when the user asks to add, fix, normalize, or improve frontmatter, properties, metadata, tags, or YAML on a note or folder of notes.

Conduct an end-of-day review to capture progress and set up tomorrow. Use when the user asks for a daily review, end of day wrap-up, EOD note, what they did today, or wants to reflect on the day and plan tomorrow.

Remove AI-generated jargon and restore human voice to text. Use when the user asks to de-AI-ify, humanize, remove AI tone, or make text sound less machine-generated.

Download files from URLs to the Obsidian attachments folder and organize them with descriptive names. Use when the user provides URLs to download, asks to save files from the web, or wants to add web content as attachments.

Work with git worktrees for isolated parallel development. Use when starting feature work in isolation, when need separate workspace without branch switching, or when cleaning up worktrees after PR merge.

Help organize and process items in the 00_Inbox folder according to the PARA method. Use when the user asks to process, clear, sort, triage, or organize their inbox.

Interactive setup wizard that helps new users create a personalized CLAUDE.md file based on their Obsidian workflow preferences. Use when the user wants to set up claudesidian, onboard a new vault, or run the bootstrap/init/setup wizard.

Install claudesidian shell command to launch Claude Code from anywhere. Use when the user wants to install a shell alias/launcher for their vault, or asks to set up the claudesidian command.

Create and edit JSON Canvas files (.canvas) with nodes, edges, groups, and connections. Use when working with .canvas files, creating visual canvases, mind maps, flowcharts, or when the user mentions Canvas files in Obsidian.

Create and edit Obsidian Bases (.base files) with views, filters, formulas, and summaries. Use when working with .base files, creating database-like views of notes, or when the user mentions Bases, table views, card views, filters, or formulas in Obsidian.

Create and edit Obsidian Flavored Markdown with wikilinks, embeds, callouts, properties, and other Obsidian-specific syntax. Use when working with .md files in Obsidian, or when the user mentions wikilinks, callouts, frontmatter, tags, embeds, or Obsidian notes.

Interactive pragmatic code review focusing on YAGNI and KISS principles. Use when the user asks for a code review, wants to check for over-engineering, or wants a YAGNI/KISS-focused review of changes.

Create a new feature branch, commit changes, push to GitHub, and open a pull request — all in one command. Use when the user asks to open a PR, create a pull request, or push changes for review.

Automatically bump version, update changelog, commit, tag, and push a new release based on recent changes. Use when the user wants to cut a release, publish a new version, bump the version, or run the release workflow.

Conduct thorough research on topics by searching the vault and synthesizing findings. Use when the user asks to research a topic, dig into something across notes, or wants a synthesis of what the vault knows about a subject.

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.

ALWAYS use before attempting any fix. Never jump to solutions - investigate root cause first. Use when encountering any technical issue, bug, test failure, or unexpected behavior.

Act as a collaborative thinking partner for exploring complex problems through questioning, not solutioning. Use when the user wants to think out loud, explore an idea, work through a problem, or asks for a thinking partner.

Intelligently upgrade claudesidian with new features while preserving user customizations using AI-powered semantic analysis. Use when the user wants to upgrade claudesidian, pull in upstream changes, or update their installation.

Create a comprehensive synthesis of the week's work and thinking. Use when the user asks for a weekly review, weekly synthesis, week-in-review, weekly wrap-up, or wants to reflect on the week and plan the next one.