Coding standards, frameworks, refactoring, and code review (18.6K skills)
Plan, scaffold, test, and deploy a code-first Deep Agent on Managed Deep Agents in LangSmith. Map requirements to supported capabilities such as tools, memory, MCP servers, schedules, and Slack, then work through the project layout and CLI deployment flow.
Set up a small LangGraph agent in a new local project using the official Python quickstart. Choose a chat model, install its provider package, configure its API key in an environment file, and run a simple example to see the agent work.
Learn how to save LangGraph state between runs, keep conversation history by thread, and share user preferences across threads. The guide also covers reviewing or replaying past states and choosing persistence settings for subgraphs.
Use this skill to add human review to LangGraph workflows. It explains how to pause with `interrupt()`, resume with `Command(resume=...)`, validate responses, and avoid duplicated side effects when a paused node runs again.
Use this skill to create and run a minimal LangChain agent locally in TypeScript, following the official quickstart. It helps you choose a model provider, set up a separate project directory, and try the weather agent example.
Add human approval before risky LangChain tool calls, resume work after approval, edits, or rejection, and create middleware for logging, retries, or other hooks. The skill also covers structured outputs with Pydantic or Zod.
Build LangChain agents with `create_agent()`, define callable tools, and control workflows with middleware. The skill also covers approval pauses, error handling, conversation state, and structured outputs that follow a defined schema.
Plan dependencies for LangChain, LangGraph, LangSmith, or Deep Agents projects in Python and TypeScript. Find core and provider packages, runtime and version requirements, integration options, and guidance for pinning versions and avoiding common compatibility problems.
Sets up a small TypeScript Deep Agent locally by following the official quickstart. It asks which model provider to use, wires in that provider’s built-in web search instead of Tavily, and runs a sample research question to show the result.
Set up and run a small LangChain agent in a fresh local Python project by following the official quickstart. Choose a model provider, add its API key locally, and see the example agent run before moving on to further customization.
Build stateful workflows and agents with LangGraph. Learn how to design state, implement nodes, connect and route edges, run or stream graphs, and handle common errors. Includes guidance on choosing the matching Python or TypeScript implementation reference.
Choose between LangChain, LangGraph, and Deep Agents for an agent project. Compare their strengths, learn how the layers can work together, set up observability variables, and identify the next skill and current documentation to consult before coding.
Choose how Deep Agents store working files and memory, from temporary thread-scoped state to storage shared across sessions or local disk access. Learn to route paths between backends and understand which data persists and where.
Build and run a minimal local Deep Agent in Python from the official quickstart. Choose a model provider and use its built-in web search for a research example, avoiding a separate search service and its API key.
Configure how Deep Agents divide work among specialized subagents, track multi-step plans, and pause sensitive tool actions for human approval. Learn when each capability fits and how to set up and resume these workflows.
Create a minimal LangGraph agent locally in TypeScript by following the official quickstart. Choose a chat model provider, set its API key in a local `.env` file, install the needed packages, and run a small calculator agent example.
Guides the use of the LangGraph CLI across an application’s lifecycle: creating a project, configuring `langgraph.json`, running locally, building or launching with Docker, and deploying to LangGraph Platform. It also covers deployment logs and common setup requirements.
Build applications with Deep Agents by configuring its agent harness, the framework layer that supplies planning, file tools, subagents, memory, approvals, and on-demand skills. Covers `create_deep_agent()`, configuration choices, and the `SKILL.md` format.
Set up Caliber to keep supported AI coding agent configuration files in sync with project changes. Check installation, install a pre-commit hook, inspect configuration freshness, and configure optional session learning or team GitHub Actions.
Use this skill to add a writer for a new AI agent platform. It guides you through defining the platform’s configuration, writing formatted files and registering the writer, then returning the paths of files created.
Use this skill to add a deterministic check that scores one aspect of agent configuration quality. It guides you through point constants, filesystem-based measurement, target filtering, score registration, and unit tests, so results remain consistent for the same project state.
Add a new command or subcommand to a Commander.js CLI. The skill covers the command handler, registration, typed option parsing, consistent error handling, and telemetry tracking, helping the new command fit the existing project patterns.
Diagnose and repair Windows Codex Desktop issues after updates, including hidden models, Fast Mode, plugins, browser use, Computer Use, and local conversation access. Choose a targeted repair and verify the relevant behavior.
Inspect a codebase to identify its development services, ports, and infrastructure needs, then prepare a single tmux-based launcher with commands for starting, stopping, checking, and managing logs. It also documents how to use the launcher.