发现 Skills

Generate comprehensive project analysis documentation for code repositories. Use when Claude needs to analyze a codebase and create detailed documentation covering project architecture, component relationships, functionality, code patterns, design decisions, and setup instructions. Uses DeepResearch multi-turn methodology for thorough analysis. Works with any local or remote repository.

Download and process videos from YouTube and other platforms. Supports video download, audio extraction, format conversion (mp4, webm), and Whisper transcription. Use when user mentions YouTube download, video conversion, audio extraction, transcription, mp4, webm, ffmpeg, yt-dlp, or whisper transcription.

Implement Convex runtime features: HTTP actions, file storage, search (full text + vector), scheduling (crons + scheduled functions), and RAG patterns. Use for webhooks, uploads, search indexes, vectorSearch actions, and background workflows. Use proactively when users mention HTTP endpoints, files, search, embeddings, or cron/schedule. Examples: - user: "Add full text search" → define searchIndex + withSearchIndex query - user: "Upload files" → generate upload URL and persist storageId - user: "Vector search" → action with ctx.vectorSearch and doc fetch - user: "Run cleanup nightly" → cronJobs + function reference

Review Vite security audit patterns for SPA and dev server security. Use for auditing VITE_* exposure, build-time secrets, and proxy configs. Use proactively when reviewing Vite apps (vite.config.ts present). Examples: - user: "Audit Vite env vars" → check for secrets with VITE_ prefix - user: "Check Vite build config" → verify define block and source maps - user: "Review Vite dev server" → check host binding and proxy security - user: "Scan Vite bundles" → search dist/ for leaked API keys or secrets - user: "Audit Vite SPA auth" → verify server-side auth vs client route guards

Partition-first log analysis methodology. Use for log searches, error analysis, pattern finding across Datadog, CloudWatch, or Kubernetes logs.

Replace text in fillable PDF forms by updating form field values. This skill should be used when users need to update names, addresses, dates, or other text in PDF form fields.

File system operations - read, write, copy, move, delete

Write a PRD optimized for AI coding agents (Claude Code, Codex). Includes standard product context plus implementation-specific sections -- file structure, code patterns, API contracts with request/response examples, test cases, migration steps, and sequenced build order. The output is designed so an AI coding tool can build directly from it.

Use when deploying or managing Kubernetes workloads that require cluster configuration, security hardening, or troubleshooting. Invoke for Helm charts, RBAC policies, NetworkPolicies, storage configuration, and performance optimization.

Use when optimizing PostgreSQL queries, configuring replication, or implementing advanced database features. Invoke for EXPLAIN analysis, JSONB operations, extension usage, VACUUM tuning, and performance monitoring.

Use when building Vue 3 applications with the Composition API, Nuxt 3, or Quasar. Invoke for Pinia, TypeScript, PWA, Capacitor mobile apps, and Vite configuration.

Enforces complete Taskwarrior integration protocol for ALL coding tasks. Activates automatically when user mentions "taskwarrior", "task warrior", "tw", or discusses task management. Decomposes all coding work into properly tracked Taskwarrior tasks with full lifecycle: task add → task start → implementation → task done. Integrates with Timewarrior for automatic time tracking.

Configure SAML 2.0 and OIDC SSO for Cursor with Okta, Microsoft Entra ID, and Google Workspace. Triggers on "cursor sso", "cursor saml", "cursor oauth", "enterprise cursor auth", "cursor okta", "cursor entra", "cursor scim".

Diagnose and fix Mistral AI common errors and exceptions. Use when encountering Mistral errors, debugging failed requests, or troubleshooting integration issues. Trigger with phrases like "mistral error", "fix mistral", "mistral not working", "debug mistral".

This skill optimizes database connection pooling for enhanced performance and resource management. It is activated when the user requests assistance with connection pooling, database performance tuning, or connection lifecycle management. Use this skill to implement and configure connection pools in various programming languages, identify optimal pool settings, and troubleshoot common connection-related issues. The skill is triggered by phrases like "connection pooling," "optimize database connections," or "improve database performance with connection pool."

通勤方案/路线规划（点到点或多点串联）。用于用户问“怎么去/怎么走/通勤方案/地铁公交换乘/打车多久多少钱/从酒店去景点或会场”等。 默认做实时查证并给来源链接与查询时间（当地时间）；不编造精确数字；写入日历必须用户明确确认。

酒店预订/选酒店/住哪里。用于用户要你“订酒店/选酒店/酒店推荐/比较官网和OTA价格/确认取消政策与早餐/靠近某地标住哪更方便”等场景。 默认官网优先、再用 OTA 比价；不自动下单；引用外部事实必须尽量给来源链接与来源当地时间；绝不编造。

Pi agent runtime integration for Cloudflare. Use when implementing the agent loop, tools, extensions, session trees, or multi-provider LLM calls. Triggers on Pi, agent runtime, tool calling, streaming, session tree, extensions, self-extending agent.

Always search before starting any work across all coding agent session histories (Claude Code, Codex, Cursor, Gemini CLI, Aider, ChatGPT) to find whatever we've discussed before.

Guide AI agents through TypeScript coding best practices including type safety, error handling, code organization, and architecture patterns. This skill should be used when generating TypeScript code, reviewing TypeScript files, creating new TypeScript modules, refactoring JavaScript to TypeScript, or when the user asks about TypeScript patterns, types, or coding standards. Keywords: typescript, types, coding standards, best practices, type safety, generics, architecture, refactoring.

Generate multiple viable solution options after research is complete, before converging on a single approach. Use when you need to explore the solution space, ask clarifying questions, and produce 3-5 distinct options to consider.

Molecular ML with diverse featurizers and pre-built datasets. Use for property prediction (ADMET, toxicity) with traditional ML or GNNs when you want extensive featurization options and MoleculeNet benchmarks. Best for quick experiments with pre-trained models, diverse molecular representations. For graph-first PyTorch workflows use torchdrug; for benchmark datasets use pytdc.

Query openFDA API for drugs, devices, adverse events, recalls, regulatory submissions (510k, PMA), substance identification (UNII), for FDA regulatory data analysis and safety research.

Use when a task is too large for a single session — provides multi-session execution with checkpoints, progress tracking, and resume capability so complex projects survive context resets and session boundaries