Skill UI

Learn how to build complex, multi-step AI agents using Cohere's Chat API v2. This skill demonstrates defining external tools (like weather APIs or database searches) and implementing the full agent loop: model calls tool -> execute tool -> feed result back -> final answer. Ideal for creating autonomous, tool-using workflows that interact with external systems.

This detailed guide outlines a production-ready, layered architecture for building complex AI applications using the Cohere API v2. It covers implementing Retrieval-Augmented Generation (RAG) pipelines, advanced tool-use agents, and multi-model orchestration. Use this structure when designing new Cohere integrations or establishing robust architectural standards for enterprise AI solutions.

This skill helps users design robust and optimal software interfaces (APIs) for a module by applying the 'Design It Twice' philosophy. It generates multiple, radically different architectural designs using parallel agents, allowing for comprehensive comparison of trade-offs (e.g., simplicity vs. flexibility, specialized vs. general-purpose). Use this when you need to explore various structural options or compare module shapes before committing to a final design.

A universal infrastructure for automatically evolving and optimizing AI agents across any domain. Using LLM-driven mutation and iterative solve-observe-evolve cycles, it enhances critical components like prompts, skills, and memory by benchmarking performance against measurable datasets, enabling the creation of self-improving AI systems.

This powerful orchestrator automates the entire GitHub issue lifecycle. It fetches issues based on criteria (labels, milestones, assignees), spawns specialized sub-agents to implement necessary code fixes, and automatically opens Pull Requests (PRs). Furthermore, it monitors open PRs, addresses review comments, and manages the remediation process, streamlining the development workflow from issue tracking to merge.

This skill orchestrates the entire development lifecycle for approved software tasks. It operates in autonomous or manual modes, rigorously applying Test-Driven Development (TDD). It automatically gathers all necessary context (specs, requirements, design, and historical learnings), validates the repository state, and dispatches specialized sub-agents to implement tasks one by one, ensuring high code quality, adherence to design, and process traceability.

This tool systematically converts a project roadmap into actionable specifications. It first analyzes features by grouping them into dependency waves. It then automatically dispatches sub-agents to generate complete feature specs (requirements, design, tasks, and JSON) for each wave. The core value lies in its comprehensive cross-spec review, which verifies data model consistency, interface alignment, and architectural integrity across the entire project, ensuring a cohesive final product.

A robust framework for building complex, long-horizon AI agents using LangGraph. It addresses common pitfalls in advanced agent design, such as unbounded state growth (Virtual FS cleanup) and subagent persona leakage (explicit message overriding). Ideal for tasks requiring multi-step planning, task delegation, and self-correction through reflection loops.

This harness provides comprehensive, reproducible evaluation pipelines for complex LLM chains and agents (LangChain/LangGraph 1.0). It integrates golden dataset management, LangSmith evaluation, RAGAS metrics, deepeval LLM-as-judge, and structured agent trajectory analysis. Use it when establishing quality benchmarks for new chains, diagnosing performance regressions after model switches, or implementing CI/CD gates to prevent quality drops.

This guide teaches how to build robust, production-grade ReAct agents using LangGraph 1.0's `create_react_agent`. It specifically addresses common agent failure modes, such as preventing infinite loops (recursion caps) and ensuring proper error propagation. Learn to define typed tools using Pydantic, manage state with checkpointers, and set precise recursion limits. Perfect for users migrating from legacy AgentExecutor or building reliable, multi-step tool-calling workflows.

A comprehensive guide to ensuring robust state management for LangGraph agents. This skill addresses common checkpointing pitfalls (like missing thread_id or non-JSON serializable state) and details best practices for implementing reliable state persistence using MemorySaver or PostgresSaver. It covers thread discipline, schema migration, and time-travel debugging for production-ready applications.

Master advanced patterns for composing LangGraph subgraphs, which is essential for building robust, multi-agent systems like planner-executors or hierarchical teams. This skill addresses critical issues such as shared state contract propagation (P21) and callback scoping (P28). Learn to manage state across independent graph components using sophisticated dispatch patterns: compiled nodes, Send fan-out, and Command bubble-up.