Data & AI

Summarize CodexBar local cost logs by model for Codex or Claude, including current or full breakdowns.

Add and live-prove a model provider with non-interactive config one-liners, without exposing credentials.

Slack archive: search, sync freshness, threads/DMs, SQL counts, and Slacrawl repo work.

Notion archive: search, sync freshness, pages/databases, Markdown exports, SQL counts, and Notcrawl repo work.

Granola archive: search, sync freshness, notes, transcripts, panels, SQL counts, and Graincrawl repo work.

Diff-scoped AI-slop cleanup pass: strip comment slop, defensive-check slop, type-laundering, and style drift from the current branch diff before autoreview.

Audit or refresh OpenClaw maturity scorecard docs from root taxonomy, maturity scores, and QA evidence artifacts without using maintainer discrawl data or committed inventory reports.

Load before calling build-workflow. Default path for all single-workflow work: new one-off workflows, existing-workflow edits, verification repairs, and workflow-local data tables. Write or edit a workspace source file, run workflow-sdk validate via workspace_execute_command, then call build-workflow with filePath. When the workflow creates or writes Data Tables, load data-table-manager first, then this skill. Do not load planning or create-tasks first. Load planning only when multiple coordinated workflows or shared cross-task data tables require a dependency-aware task graph.

ONLY for coordinated multi-artifact work: multiple workflows with dependencies, shared data-table schema/migration across tasks, or the user explicitly asked to review a plan first. Load create-tasks via load_tool before calling it (search "create tasks" if not visible). Do NOT use for new one-off workflows, single-workflow edits, verification-only requests, or standalone data-table ops — use workflow-builder or data-table-manager instead.

Handles system follow-up turns: planned-task-follow-up (synthesize, replan, build-workflow, checkpoint), background-task-completed, running-tasks context, and create-tasks silence rules. Load whenever any of these tags appear or after calling create-tasks.

Classifies automation requests using two decisions: anchor (which primitive owns the top-level control flow — workflow-anchored, agent-anchored, needs-clarification, or out-of-scope) and embeds_other (whether the other primitive appears embedded inside — an agent step inside a workflow, or a workflow invoked as an agent tool). Must be used whenever the current turn requires choosing or reconsidering the intent of an automation request, including compound requests, independent automations introduced mid-build, one-off questions or reports that need external systems you cannot query directly, and requests that need clarification before an anchor can be chosen. Do not load for routine edits or extensions when the conversation already targets a workflow or Agent.

Load before calling data-tables or parse-file. Use for natural standalone requests like "what data tables do I have?", "show/list my tables", or "what columns are in this table?", and whenever the user asks to list, show, create, inspect, import, seed, query, update, clean up, rename columns in, or delete data tables and rows, especially from CSV/XLSX/JSON attachments. Also load before building or planning workflows that create or write to Data Tables (then load workflow-builder before build-workflow).

Builds and maintains configuration-based evaluations on a workflow with the eval-config tool. Use when the user asks to set up, add, view, change, or remove an evaluation, score, grade, or judge a workflow's output, or measure answer quality against a test dataset. This is the only eval form Instance AI handles — it does not touch on-canvas evaluation nodes.

Authors a new Instance AI workflow eval case — written locally as JSON, calibrated against a real build, then pushed to the LangTracer suite CI runs — build cases, behaviour/process cases, credential cases, and seeded (mid-conversation) cases — with intent-driven expectations. Use when adding or changing an Instance AI workflow eval, or debugging why one is flaky.

苹果FoundationModels框架用于设备上的LLM——文本生成、使用@Generable进行引导生成、工具调用，以及在iOS 26+中的快照流。

通过 fal.ai MCP 实现统一的媒体生成——图像、视频和音频。涵盖文本到图像（Nano Banana）、文本/图像到视频（Seedance、Kling、Veo 3）、文本到语音（CSM-1B），以及视频到音频（ThinkSound）。当用户想要使用 AI 生成图像、视频或音频时使用。

通过Exa MCP进行神经搜索，适用于网络、代码和公司研究。当用户需要网络搜索、代码示例、公司情报、人员查找，或使用Exa神经搜索引擎进行AI驱动的深度研究时使用。

构建一个全自动化的AI驱动数据收集代理，适用于任何公共来源——招聘网站、价格信息、新闻、GitHub、体育赛事等任何内容。按计划进行抓取，使用免费LLM（Gemini Flash）丰富数据，将结果存储在Notion/Sheets/Supabase中，并从用户反馈中学习。完全免费在GitHub Actions上运行。适用于用户希望自动监控、收集或跟踪任何公共数据的场景。

LLM API 使用成本优化模式 —— 基于任务复杂度的模型路由、预算跟踪、重试逻辑和提示缓存。

团队中人工智能代理生成大部分实施输出的工程运营模型。

针对AI代理故障的结构化自调试工作流程，包括捕获、诊断、受限恢复和内省报告。

设计和优化AI代理的动作空间、工具定义和观察格式，以提高完成率。

実写素材のカット、構築、強化のためのAI支援ビデオ編集ワークフロー。生の撮影素材からFFmpeg、Remotion、ElevenLabs、fal.aiを経て、DescriptまたはCapCutで最終仕上げを行う完全なパイプラインをカバーする。ユーザーがビデオの編集、素材のカット、vlogの作成、またはビデオコンテンツの構築を望む場合に使用する。

構造化テキストの解析に正規表現と大規模言語モデルのどちらを使うかを選択するための意思決定フレームワーク——まず正規表達式から始め、信頼度の低いエッジケースにのみ大規模言語モデルを追加する。

日本語翻訳：このファイルは prompt-optimizer 用の日本語翻訳が必要です

日本語翻訳：このファイルは llm-trading-agent-security 用の日本語翻訳が必要です

ヘルスケアAIモデル評価ハーネス、臨床メトリクス、およびレギュレーション遵守の検証。

fal.ai MCPによる統合メディア生成（画像、動画、音声）。テキストから画像（Nano Banana）、テキスト/画像から動画（Seedance、Kling、Veo 3）、テキストから音声（CSM-1B）、動画から音声（ThinkSound）をカバーします。ユーザーがAIで画像、動画、音声を生成したい場合に使用します。

Exa MCPによるウェブ、コード、企業調査のためのニューラル検索。ユーザーがウェブ検索、コード例、企業情報、人物検索、またはExaのニューラル検索エンジンを使ったAI駆動の詳細調査を必要とする場合に使用します。

複数のAIエージェントとタスク集約ワークフローを調整します。複数のワーカーで作業を分配し、エラーを処理し、結果をマージ。

任意のパブリックソース（ジョブボード、価格、ニュース、GitHub、スポーツなど）用の完全自動化されたAI搭載データ収集エージェントを構築します。スケジュールでスクレイプし、無料LLM（Gemini Flash）でデータを豊かにし、Notion/Sheets/Supabaseに結果を保存し、ユーザーフィードバックから学習します。GitHub Actions上で100％無料で実行。ユーザーがパブリックデータを自動的に監視、収集、または追跡したい場合に使用します。

LLM APIの使用量のコスト最適化パターン — タスクの複雑さによるモデルルーティング、予算追跡、リトライロジック、プロンプトキャッシング。

ClickHouse database patterns, query optimization, analytics, and data engineering best practices for high-performance analytical workloads.

AI エージェントが大量の実装出力を生成するチームのためのエンジニアリング運用モデル。

タスクごとのバジェット、支出コントロール、ノンカストディアルウォレットを備えた x402 決済実行を AI エージェントに追加します。agentwallet-sdk を通じて Base をサポートし、OKX Payments / OKX エージェント決済プロトコルを通じて X Layer をサポートします。

キャプチャ、診断、封じ込め回復、内省レポートを使用した AI エージェント障害のための構造化された自己デバッグワークフロー。

AI エージェントのアクション空間、ツール定義、観測フォーマットを設計・最適化して完了率を向上させます。

Thread-safe data persistence in Swift using actors — in-memory cache with file-backed storage, eliminating data races by design.

PyTorch deep learning patterns and best practices for building robust, efficient, and reproducible training pipelines, model architectures, and data loading.

PostgreSQL database patterns for query optimization, schema design, indexing, and security. Quick reference for common patterns, index types, data types, and anti-pattern detection. Based on Supabase best practices.

Production machine-learning engineering workflow for data contracts, reproducible training, model evaluation, deployment, monitoring, and rollback. Use when building, reviewing, or hardening ML systems beyond one-off notebooks.

Database migration best practices for schema changes, data migrations, rollbacks, and zero-downtime deployments across PostgreSQL, MySQL, and common ORMs (Prisma, Drizzle, Django, TypeORM, golang-migrate). Use when planning or implementing database schema changes.

Operate as an agentic engineer using eval-first execution, decomposition, and cost-aware model routing. Use when AI agents perform most implementation work and humans enforce quality and risk controls.

AI-assisted video editing workflows for cutting, structuring, and augmenting real footage. Covers the full pipeline from raw capture through FFmpeg, Remotion, ElevenLabs, fal.ai, and final polish in Descript or CapCut. Use when the user wants to edit video, cut footage, create vlogs, or build video content.

Production machine-learning engineering workflow for data contracts, reproducible training, model evaluation, deployment, monitoring, and rollback. Use when building, reviewing, or hardening ML systems beyond one-off notebooks.

Unified media generation via fal.ai MCP — image, video, and audio. Covers text-to-image (Nano Banana), text/image-to-video (Seedance, Kling, Veo 3), text-to-speech (CSM-1B), and video-to-audio (ThinkSound). Use when the user wants to generate images, videos, or audio with AI.

Structured self-debugging workflow for AI agent failures using capture, diagnosis, contained recovery, and introspection reports. Use when an agent run fails and you need a reproducible diagnosis instead of a retry.

Query PyTorch CI, GitHub Actions, HUD, Grafana, and infrastructure metrics. Use when users ask about CI duration, job failures, queue times, workflow trends, runner health, dashboard data, or PyTorch infrastructure metrics.

Reference for the Claude API / Anthropic SDK — model ids, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration. TRIGGER — read BEFORE opening the target file; don't skip because it "looks like a one-liner" — whenever: the prompt names Claude/Anthropic in any form (Claude, Anthropic, Fable, Opus, Sonnet, Haiku, `anthropic`, `@anthropic-ai`, `claude-*`, `us.anthropic.*`, `[1m]`); the user asks about an LLM (pricing/model choice/limits/caching) — never answer from memory; OR the task is LLM-shaped with provider unstated (agent/MCP/tool-definition/multi-agent/RAG/LLM-judge/computer-use; generate/summarize/extract/classify/rewrite/converse over NL; debugging refusals/cutoffs/streaming/tool-calls/tokens). SKIP only when another provider is being worked on (overrides all triggers): OpenAI/GPT/Gemini/Llama/Mistral/Cohere/Ollama named in the query; OR `grep -rE 'openai|langchain_openai|google.generativeai|genai|mistralai|cohere|ollama'` over the project hits (run this grep FIRST if no provider named — don't Read the file).

Build and sharpen a project's domain model. Use when discussing codebase terminology, writing or editing a CONTEXT.md, or recording or editing an ADR.