Complex coding tasks require multiple iterations and coordinated changes across files. This skill spawns autonomous agents that work through a structured plan with built-in review gates. The swarm completes multi-step development work without constant back-and-forth interaction.

An AI skill for Claude, Codex, and Claude Code

Browse and interact with ClawDirect, a directory of social web experiences for AI agents. Get MCP tool usage instructions and API documentation. Connect via ATXP authentication.

An AI skill for Claude, Codex, and Claude Code

Manage Things 3 via the things CLI on macOS — add/update projects and todos via URL scheme, read/search/list from local Things database. Use when a user asks to add a task to Things, list inbox/today/upcoming, search tasks, or inspect projects/areas/tags.

Manage Apple Reminders via remindctl CLI — list, add, edit, complete, delete. Supports lists, date filters, and JSON/plain output. Syncs to iPhone/iPad Reminders app.

Manage Apple Notes via the memo CLI on macOS — create, view, edit, delete, search, move, and export notes. Use when a user asks to add a note, list notes, search notes, or manage note folders.

An AI skill for Claude, Codex, and Claude Code

Building LLM applications without proper architecture leads to wasted time and resources. This skill provides battle-tested patterns from production systems like Karpathy's HN Time Capsule and Vercel's d0 agent.

Single-agent systems face context limits, sequential processing bottlenecks, and inflexible tool configurations. This skill provides patterns for distributing work across multiple AI agents with explicit coordination, handoff protocols, and failure handling.

An AI skill for Claude, Codex, and Claude Code

Measuring agent quality requires structured approaches. This skill provides multi-dimensional evaluation rubrics, test set management, and production monitoring to continuously assess and improve agent performance.

Limited context windows constrain complex tasks and increase API costs. This skill provides proven techniques to double or triple effective context capacity through compaction, observation masking, KV-cache optimization, and context partitioning.

Context is the foundation of effective AI agent performance. This skill teaches you how to understand, design, and optimize context in agent systems. Learn the principles of progressive disclosure, attention mechanics, and context budgeting to build more capable AI agents.

AI agents lose performance as conversations grow longer. This skill diagnoses lost-in-middle problems, context poisoning, and attention failures to restore optimal agent performance. It provides patterns for recognizing and mitigating context degradation.

Long agent sessions generate millions of tokens that exceed context limits. This skill provides structured compression strategies that reduce token usage while preserving file paths, error messages, and decisions so agents can continue working without re-fetching information.

Transform RDF data into cognitive agent architectures using formal Belief-Desire-Intention patterns. Enable AI systems to model rational agency with traceable reasoning chains and semantic interoperability for multi-agent systems.

Manual evaluation of LLM outputs is slow and inconsistent across reviewers. This skill provides production-grade techniques for automated evaluation using LLMs as judges. It includes direct scoring rubrics, pairwise comparison with position bias mitigation, and confidence calibration.

Creating Solana blockchain projects from scratch requires understanding Anchor framework conventions, Rust smart contract structure, and Next.js frontend integration. This skill generates complete production-ready projects with proper structure, configuration, and wallet integration.

Simplify AI model development and deployment on Huawei Cloud ModelArts. This skill provides unified resource management for training jobs, models, inference services, and notebooks with built-in security features.