3,313 件 ・ 関連度順
概要と使いどころ
Audit an LLM eval pipeline and surface problems: missing error analysis, unvalidated judges, vanity metrics, etc. Use when inheriting an eval system, when unsure whether evals are trustworthy, or as a starting point when no eval infrastructure exists. Do NOT use when the goal is to build a new evaluator from scratch (use error-discovery, write-judge-prompt, or validate-evaluator instead).
日本語の概要は準備中です。原文の説明を表示しています。
Design LLM-as-Judge evaluators for subjective criteria that code-based checks cannot handle. Use when a failure mode requires interpretation (tone, faithfulness, relevance, completeness). Do NOT use when the failure mode can be checked with code (regex, schema validation, execution tests); use `write-code-eval`. To validate an existing judge, use `validate-evaluator`.
日本語の概要は準備中です。原文の説明を表示しています。
Write code evaluators for known failure modes with objective rules. Use when code can check the rule from a trace, with or without a reference answer. Use `write-judge-prompt` when the rule requires interpretation.
日本語の概要は準備中です。原文の説明を表示しています。
Diagnoses and repairs macOS TCC dialogs and silent denials: Full Disk Access, Screen Recording, Microphone, Camera, Accessibility, Automation. Use for recurring prompts, LaunchAgent/uv jobs that fail only in background, broken grants, or automating Full Disk Access (权限弹窗 / 授权了没用 / 完全磁盘访问). Checks requester and grants first. Not for launchd design (macos-watchdog) or app permission UX (developing-ios-apps).
日本語の概要は準備中です。原文の説明を表示しています。
MUST be used right after changing any CLI color config (delta/bat/themes/lazygit pager) — renders the terminal's colored output to a PNG so Claude can actually SEE contrast, alignment, and highlighting instead of raw ANSI codes. Use for delta git diff colors, bat syntax highlighting, starship prompt, eza/ls colors, or ripgrep matches. Trigger phrases: 看终端效果, 终端截图, 验证配色, terminal screenshot.
日本語の概要は準備中です。原文の説明を表示しています。
Sets up QA testing processes: test strategies, Google Testing Standards test cases, P0-P4 bug tracking, quality metrics, progress reports, autonomous execution via master prompts, and OWASP security testing, with third-party QA handoff docs. Use when establishing QA infrastructure, writing test cases, executing test plans, or tracking bugs. Not for auditing an already-rendered UI (use frontend-visual-qa).
日本語の概要は準備中です。原文の説明を表示しています。
Tests/benchmarks a third-party LLM endpoint (OpenAI- or Anthropic-compatible): availability, fidelity, speed, concurrency, protocol compliance, quality regression. Use before hand-rolling a curl loop, when onboarding a provider, or debugging "system prompt 不生效" / a tok/s claim — 测评/压测一个模型或渠道. Not for TTS/voice-clone supplier eval (use the audio skill).
日本語の概要は準備中です。原文の説明を表示しています。
Extract design systems from reference UI images and generate implementation-ready UI design prompts. Use when users provide UI screenshots/mockups and want to create consistent designs, generate design systems, or build MVP UIs matching reference aesthetics.
日本語の概要は準備中です。原文の説明を表示しています。
Use when asked 帮我写教案, 写一份新课标教案, 教学设计, 单元教学设计, 大单元教案, or a lesson plan in the format Chinese schools expect. Produces a complete 教案 in Chinese with 教材分析, 学情分析, 教学目标 tied to the subject's 核心素养, 教学重难点, 教学方法, a timed 教学过程 with teacher and student activity, 板书设计, layered 作业设计 within the 双减 time limits, and 教学反思 prompts, with curriculum references marked to confirm against the current 课程标准.
日本語の概要は準備中です。原文の説明を表示しています。
Get started birding from where you are — what you're likely to see, how to tell confusing species apart, and a simple life-list to track sightings. Use when asked to start birdwatching, what bird did I see, help me identify a bird, or set up a birding log. Produces a likely-species list for your area and season, ID prompts (the field marks and sounds that separate look-alikes), a beginner gear-and-timing note, and a lightweight life-list format — pointing you to a live ID app to confirm any specific sighting.
日本語の概要は準備中です。原文の説明を表示しています。
Design an evaluation plan for an LLM or AI feature before shipping it. Use when asked how to evaluate a prompt/model/agent, set up an eval harness, define quality metrics for an AI feature, or build a regression gate. Produces an eval plan — task definition, datasets, metrics & rubrics, baselines, automated + human evals, a pass bar, and a regression gate.
日本語の概要は準備中です。原文の説明を表示しています。
Build the context an AI needs to do a task well — the background, constraints, examples, and format it can't guess — so you get a great result on the first try instead of a generic one you have to keep correcting. Use when asked why does AI give me generic answers, how do I give AI better context, my AI results are mediocre, or how do I get it right the first time. Produces the specific context this task needs (who/what/constraints/examples/format), a reusable primer you can paste ahead of the request, the difference between a starved prompt and a well-briefed one, and what to leave out — turning vague back-and-forth into a strong first result.
日本語の概要は準備中です。原文の説明を表示しています。
Run a blameless postmortem for an incident caused by an AI agent or LLM feature — hallucinated facts shipped to users, runaway tool use, prompt injection, cost blowouts, or wrong actions taken autonomously. Use when asked to write up an AI incident, analyse why an agent did something wrong, or produce corrective actions after an LLM failure. Produces a structured postmortem with trace reconstruction, a root-cause layer analysis, and corrective actions including a permanent regression case. For non-AI production incidents use incident-postmortem.
日本語の概要は準備中です。原文の説明を表示しています。
Design an AI-assisted workflow for a recurring task — which steps to hand to AI, which to keep human, and how they connect — so you get leverage without losing quality or control. Use when asked how do I use AI for [process], automate this with AI, design an AI workflow, or where does AI fit in my process. Produces a map of the task's steps split into AI-does / human-does / human-checks, the right tool/prompt for each AI step, the hand-offs and review points, the failure modes to guard against, and a start-small rollout — turning a manual process into a reliable AI-assisted one that keeps you in control.
日本語の概要は準備中です。原文の説明を表示しています。
Use this skill when planning, redesigning, or reviewing Xiaohongshu carousel visuals, covers, 3:4 image posts, page-by-page visual direction, image-generation prompts, style selection, layout critique, or XHS captions. It acts as a senior visual director for 小红书图文, not a generic copywriting assistant.
日本語の概要は準備中です。原文の説明を表示しています。
Read and write Langfuse: browse traces and observations, list prompts and datasets, score traces, add dataset items. Trigger phrases: langfuse, llm observability.
日本語の概要は準備中です。原文の説明を表示しています。
HeyGen avatar and talking-head video: prompt-to-video agent, multi-scene avatar video, status polling, avatar and voice lists. Trigger phrases: heygen, avatar video, talking head video, video agent.
日本語の概要は準備中です。原文の説明を表示しています。
ALWAYS invoke this skill when you need the user to act - run a command, paste a secret, click, approve - and whenever they ask how to do something or say they do not know what to do: "step by step", "walk me through it", "what do I do", "what should I do", "I don't understand what to do", "explain what I need to do", in any language. Picking which task comes next is os-whats-next; this skill is for doing the thing in front of you. First earn the ask: try it yourself, find another route, shrink it to the part only they can do. Then one action per step, commands labelled by what they touch, no jargon. Commands are single lines that prompt for any value - typing hidden for secrets - and confirm in plain words. Afterwards verify their step.
日本語の概要は準備中です。原文の説明を表示しています。
Run fast, progressive experiments to map tunable parameters and their effects. Use when optimizing code, prompts, configurations, or other artifacts against a goal or benchmark, or when a long research run needs a planned hypothesis queue, focused trials, and durable findings.
日本語の概要は準備中です。原文の説明を表示しています。
Use the local Codex CLI as an independent second agent. Two branches — (1) proactively run `codex review` for a second opinion after completing a substantive change, before presenting it as done or committing; (2) delegate a well-defined implementation task via `codex exec`, ONLY when the user explicitly asks for Codex to do it. Also covers how to prompt Codex.
日本語の概要は準備中です。原文の説明を表示しています。
Anthropic Claude API patterns for Python and TypeScript. Covers Messages API, streaming, tool use, vision, extended thinking, batches, prompt caching, and Claude Agent SDK. Use when building applications with the Claude API or Anthropic SDKs.
日本語の概要は準備中です。原文の説明を表示しています。
交互式 AI Agent 开发工作坊：通过 6 阶段深度协作对话，引导用户完成 Agent 需求分析、架构设计、 工具定义、Prompt 与编排设计、代码生成、验证迭代，产出可直接运行的 Agent 项目。 框架无关设计优先，支持 LangChain / EINO / AutoGen / AgentScope / CrewAI 等 ADK 框架。
日本語の概要は準備中です。原文の説明を表示しています。
Review code for security problems; scan for vulnerabilities, secrets, dependency and prompt risks. Use when: asked whether code is safe to ship, even one small handler.
日本語の概要は準備中です。原文の説明を表示しています。
Run one prompt through headless Claude with scoped permissions and a time bound. Use when: scripting or automating a `claude -p` call, even a simple one.
日本語の概要は準備中です。原文の説明を表示しています。