DeepSeek Skill Leaderboard

Converts code, docs, and media into a persistent knowledge graph with community detection for architectural analysis and GraphRAG integration.

A frontend design skill for auditing UX, optimizing visual hierarchy, and implementing advanced UI effects and reusable design tokens.

Security workflows for AI agent prompt injection, APK hooking, and Active Directory certificate abuse within sandbox environments.

Implements persistent state management for AI agents using local markdown files to track task plans, findings, and session progress.

Creates reusable Person Profiles from source material about colleagues, relationships, or celebrities.

A research automation toolkit for searching arXiv, summarizing academic papers, analyzing ML experiment results, and planning ablation studies.

Generates and optimizes tool-specific prompts for LLMs, coding agents, and generative AI tools.

A library of agent skill files covering the AI research lifecycle: autonomous experimentation, model architecture, fine-tuning, mechanistic interpretability, distributed training, safety, optimization, evaluation, inference serving, and MLOps.

Converts IaC, SQL, and API schemas into editable draw.io diagrams, supporting C4, SysML, UML, and ERD standards.

A CLI tool to create, read, and edit Office documents locally, supporting conversions between PDF, DOCX, XLSX, PPTX, Markdown, HTML, and CSV.

Analyzes reference images to extract high-fidelity prompts for recreating visual styles, compositions, and textures in AI image generators.

Converts images into structured JSON evidence including OCR and layout semantics for text-only AI agents.

A writing toolkit for fiction that provides specific skills for prose review, minimal refactoring, full rewrites, and Hemingway-style stylistic editing.

A vision helper that uses node.js to convert images from files, URLs, or the clipboard into text descriptions.

Analyzes existing articles to extract writing style DNA and removes AI-generated linguistic patterns using a rule-based whitelist.

Generates animated architecture and data-flow diagrams from codebases or pull requests to visualize system changes.

Developer skills for AI systems, data pipelines, APIs, architecture, accessibility, security, performance, and framework migration.

An architecture audit and technical debt assessment tool with automated remediation capabilities for codebase maintenance.

A management suite for creating, reviewing, and governing custom Claude Code skills and session configurations.

A set of 28 critical-thinking skills and mental models, including Bounded Rationality and Circle of Competence, for Agent Skills-compatible AI tools.

A collection of skills for AI agents to perform architecture-aware brainstorming, plan execution, and anti-entropy governance in complex codebases.

Dubs and subtitles video files into 33 languages with style-appropriate voices for multilingual content production.

Provides the technical specifications and file checklists for building channel extensions, transport layers, and webhook guards within Tel-Agent.

A marketing skill set for Google Search Console analysis, SEO monitoring, and tracking AI-driven search traffic.

A tool for auditing and improving website citability for AI search engines using a 47-method engine, robots.txt configuration, and llms.txt generation.

Agent orchestration skills for browser and desktop GUI automation and transcript compaction for context window optimization.

A design kit for creating manufacture-realistic parametric CAD models through a code-based workflow, including tools for design specifications and model validation.

Academic research tools for scientific paper searching, structured data extraction from full-text studies, and citation management.

Covers idiomatic Synalinks development: data models, program APIs, modules, control flow, agents, RAG, training, rewards, and providers.

Provides PR-driven history and optimization evidence for SGLang, vLLM, TensorRT-LLM, and TokenSpeed serving frameworks.

Optimizes configurable system parameters for model training, inference, and black-box scientific evaluations using scalar objectives.

Performs exhaustive, time-windowed literature searches to produce DOI and reference lists with measurable recall for systematic reviews.

A suite of tools for GEO analysis and content optimization, featuring a ChatGPT crawler that aggregates search data into JSON and HTML reports.

Tools for generating platform-compliant ad copy and designing statistically rigorous A/B test plans with sample size calculations.

Management utilities for Codex that include model bridge auditing, task handoff continuity, and enterprise clone building.

An orchestration skill for multi-file builds and refactors that separates high-level planning from delegated implementation.

Generates on-brand HTML presentation decks and responsive websites using brand DNA extraction and codified design rules for slides and web.

A design tool for creating frontend UI, multi-page flows, and graphics using AI models on a visual canvas.

An OSINT and CTI analysis toolkit for infrastructure pivoting, breach triage, digital footprint review, and structured threat reporting.

A library of business skills covering the 12-phase dropshipping pipeline, pricing strategies, and psychology-based marketing optimization.

Adds web search, X (Twitter) search, and URL fetching capabilities to AI models via a CLI-based search bridge.

A directory for discovering and comparing Jev and TypeSafe SDKs, integrations, and evaluation resources.

A routing mechanism that transforms underspecified requests into typed IntentSpecs to ensure agent actions align with verified user intent.

Generates cinematic AI video prompts using a five-stage structure for narrative shorts and complex visual sequences.

An iOS App Store Optimization tool for keyword research, metadata localization in 50 locales, and territory pricing.

Generates valid ComfyUI workflow JSON from natural language, with 34 templates for image, video, audio, and 3D, plus LLM integration for multi-stage pipelines.

A routing framework to reduce operational costs by assigning routine agent tasks to low-cost models and complex reasoning to premium tiers.

Generates and validates synthesizable Verilog-2001 RTL, providing tools for testbench planning, semantic annotation, and Vivado/xsim validation.

An end-to-end training assistant for Unsloth that optimizes memory, supports SFT, DPO, GRPO, and other RL methods, and exports to local or cloud deployment targets.

Extracts structured trading strategies from finance documents, generates validated Backtrader code, runs backtests, and compares results against published metrics.