Single AI agents hit context limits when handling complex tasks. This skill teaches proven patterns to distribute work across multiple agents while maintaining coordination and avoiding common failure modes.

Agents lose all knowledge when sessions end. This skill provides patterns for layered memory architectures that preserve entity knowledge, track relationships, and enable temporal queries across sessions.

An AI skill for Claude, Codex, and Claude Code

Deploying untested code to production often causes bugs, downtime, and user-facing errors. This skill enforces a mandatory verification workflow that requires successful build completion, local server startup, browser testing, and passing tests before any deployment command can run.

LLM context windows limit task complexity and increase costs as conversations grow. This skill provides techniques to extend effective capacity through strategic compression, observation masking, KV-cache optimization, and context partitioning.

Understanding context windows and attention mechanics is crucial for building effective AI agents. This skill teaches you how to optimize context usage, manage token budgets, and design efficient agent architectures.

LLM agents fail unpredictably as context grows. This skill helps you identify and fix context degradation issues like lost-in-middle effects, poisoning, and attention collapse so your agents stay reliable at scale.

Long AI agent sessions quickly exceed context windows, causing failures. This skill provides proven compression strategies that preserve critical information while reducing tokens by 98%+.

Building reliable LLM evaluation systems is challenging due to systematic biases. This skill provides production-ready patterns for direct scoring, pairwise comparison, and bias mitigation to create trustworthy automated assessment pipelines.

Technical teams often lose context when decisions are made informally. This skill creates structured Architecture Decision Records that capture the problem, constraints, options considered, and consequences for future reference.

Struggling to write concise daily standups that actually help your team? This skill transforms your work activity into properly formatted standup reports that focus on outcomes, not just activities.

Creating your first Claude Code skill can be confusing without guidance. This template demonstrates the correct structure and principles for building effective skills.

Building domain expertise skills from scratch is challenging without a clear structure. This template provides a ready-made framework with principles, processes, and output templates so you can focus on your domain knowledge instead of format.

Code reviews often lack structure, leading to inconsistent feedback and missed issues. This skill provides a systematic 3-pass review method with clear guidelines for giving constructive feedback.

Manually writing API tests for complex endpoints with dependencies is time-consuming and error-prone. This skill automatically generates executable Python test scripts from OpenAPI or GraphQL specifications that run endpoints in the correct dependency order.

Struggling to coordinate work between Cursor and Claude Code? This skill provides clear guidance on the 2-agent workflow, helping PMs and workers collaborate effectively. Learn the exact commands, task management system, and handoff procedures for seamless project execution.

Managing handoffs between PM and implementation roles requires consistent templates and clear communication. This skill provides structured handoff templates for task delegation and automated fix workflows for review comments.

Non-technical users often feel stuck during software development. This skill provides guided responses to natural language questions like 'what should I do?' or 'how do I proceed?', helping VibeCoders navigate the development workflow.

Failed builds and broken tests waste development time. This skill verifies code builds, recovers from common errors, and applies approved review fixes while enforcing quality standards that prevent test tampering.

Stop manually coding UI components from scratch. This skill generates accessible, responsive UI components and feedback forms using shadcn/ui and modern best practices.