Devin Skill Leaderboard

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

Audits repositories for over-engineering and aggregates shortcut comments into a technical debt ledger.

A collection of token optimization skills that use a compressed communication pattern to reduce LLM context consumption.

A large-scale collection of agentic skills and templates for building specialized AI automation capabilities.

A collection of 817 structured cybersecurity skills for AI agents, covering red-teaming techniques and credential access mapped to MITRE ATT&CK and NIST CSF.

Provides file-based persistence for AI agents using task_plan.md, findings.md, and progress.md to track multi-step projects across session resets.

Produces tool-specific, copyable prompt blocks from a rough intent, with routing, intent extraction, and zero wasted tokens.

A set of agent orchestration skills covering self-scheduling, multi-agent delegation loops, and direct browser automation via CDP.

A collection of skills for empirical research, covering causal inference, econometrics, and the end-to-end academic writing process.

Product management skills for AI strategy, acquisition channel optimization, and AI-assisted prototyping to validate ideas without engineering bottlenecks.

AI rules for Flutter development focusing on feature-first architecture, Bloc/Cubit state management, and WCAG accessibility compliance.

Agent skill for the Bitget Wallet API: token security checks, market data, swap quotes, RWA stock trading, address discovery, K-line charts, and wallet operations across Ethereum, Solana, BNB Chain, Base, Arbitrum, Polygon, Morph, and Tron.

Provides intent-to-tool routing, safe defaults, and workflow patterns for Ai2 Asta MCP tools covering academic paper search, citation traversal, and author discovery.

A developer toolkit for deploying isolated code sandboxes, durable orchestration, and versioned cloud storage for AI agents.

Captures and recovers architectural decisions, rejected alternatives, and historical context to maintain codebase reasoning for humans and AI agents.

An agent orchestration tool that offloads repetitive tasks to cheaper models based on live benchmarks and token limits.