langchain-ai/deepagents
17 skills19.7k installs465k starsGitHub
Install
npx skills add https://github.com/langchain-ai/deepagentsSkills in this repo
1Langgraph DocsLangGraph documentation enables developers to build stateful agents and multi-agent systems using graph-based state machines. It provides essential patterns for agent orchestration, including cyclical workflows where agents loop through decision steps, human approval gates that pause execution for review, and coordination mechanisms between multiple autonomous agents. Developers use LangGraph when implementing complex agentic systems beyond simple prompt-response chains, requiring state persistence across agent steps, conditional routing between agent nodes, and human feedback integration into agent loops. Key developer workflows include defining agent state schemas, constructing directed graphs of agent operations, persisting conversation memory across interactions, and implementing execution breakpoints for human review before autonomous actions execute. Build agents as directed graphs with nodes representing discrete operations and edges controlling state transitions Define persistent state machines that maintain context across multi-step agent interactions and human interventions Implement human-in-the-loop checkpoints where agents pause execution pending human approval or fee.4.3kinstalls2Web Researchweb-research orchestrates multi-source online research using delegated subagents and file-based handoffs. Before delegating, the agent must create a research folder, break the question into two to five non-overlapping subtopics, and write research_plan.md with the main question, subtopics, expected information, and synthesis plan. Simple fact-finding uses one to two subtopics; comparative analysis assigns one subtopic per comparison element up to three; complex investigations use three to five. For each subtopic, the task tool spawns a research subagent with a clear question, instructions to save findings to research_[topic]/findings_[subtopic].md, and a budget of three to five web searches. Up to three subagents may run in parallel. After completion, list_files and read_file review local findings, then synthesize a response with integrated insights and source URLs; optional research_report.md captures the final write-up. fetch_url handles remote URLs; read_file is for local files only. Best practices emphasize planning before delegating, distinct subagent scopes, file-based communication, and stopping after sufficient coverage rather than over-searching.3.6kinstalls3Social MediaBefore writing any social media content you MUST delegate research 1 Use the task tool with subagent_type researcher 2 In the description specify BOTH the topic AND where to save task subagent_type researcher description Research TOPIC Save findings to research slug md Example task subagent_type researcher description Research renewable energy trends in 2025 Save findings to research renewable energy md The social media skill documents workflows prerequisites and usage patterns grounded in its repository SKILL md Agents should follow the documented steps respect safety and permission notes and cite only capabilities described in the source It triggers on phrases matching the skill description and integrates with the agent toolchain for the tasks outlined in the documentation2.4kinstalls4Arxiv SearchThe arxiv-search skill. Searches arXiv for preprints and academic papers, retrieves abstracts, and filters by topic. Use when the user asks to find research papers, search arXiv, look up preprints, find academic articles in physics, math, CS, biology, statistics, or related fields. The workflow follows the source SKILL.md contract with progressive reference loading, clear trigger phrases, and practical steps developers can apply directly in agent sessions. The workflow follows the source SKILL.md contract with progressive reference loading, clear trigger phrases, and practical steps developers can apply directly in agent sessions. The workflow follows the source SKILL.md contract with progressive reference loading, clear trigger phrases, and practical steps developers can apply directly in agent sessions. The workflow follows the source SKILL.md contract with progressive reference loading, clear trigger phrases, and practical steps developers can apply directly in agent sessions.2.1kinstalls5Blog PostThe blog-post skill. Writes and structures long-form blog posts, creates tutorial outlines, and optimizes content for SEO with cover image generation. Use when the user asks to write a blog post, article, how-to guide, tutorial, technical writeup, thought leadership piece, or long-form content. In the description, specify BOTH the topic AND where to save: Example: 3. After research completes, read the findings file before writing ## Output Structure (Required) **Every blog post MUST have both a post AND a cover image:** Example: A post about "AI Agents in 2025" → **You MUST complete both steps:** 1. Generate a cover image using and save to **A blog post is NOT complete without its cover image.** ## Blog Post Structure Every blog post should follow this structure: ### 1. Hook (Opening) - Start with a compelling question, statistic, or statement - Make the reader want to continue - Keep it to 2-3 sentences ### 2.2kinstalls6Skill Creatorskill-creator from langchain-ai/deepagents is the canonical guide for building effective agent skills. It walks through creating, initializing, scaffolding, updating, and validating skills that extend Claude, Cursor, or other agents with specialized knowledge, workflows, or integrations. Developers reach for skill-creator on prompts like create a skill, new skill, make a skill, skill for X, or how do I create a skill. The skill explains location conventions, structure, design patterns, and validation so new capabilities remain discoverable and maintainable across agent environments.1.2kinstalls7Query Writingquery-writing is a LangChain DeepAgents skill that converts natural-language data requests into correct, executable SQL. For simple questions the workflow identifies the target table, loads column definitions with sql_db_schema, drafts SELECT statements with WHERE, LIMIT, and ORDER BY, then runs sql_db_query and formats results. Complex requests add multi-table JOINs, aggregations, and subqueries. Developers reach for query-writing when a coding agent must answer ad-hoc reporting questions, prototype analytics queries, or debug data retrieval without hand-writing every statement against unfamiliar schemas.1.1kinstalls8Schema Explorationschema-exploration is a Deep Agents skill that standardizes read-only database reconnaissance before any query or feature code ships. Step one calls sql_db_list_tables to return every available table. Step two invokes sql_db_schema with selected table names to surface column names, data types such as INTEGER, TEXT, and DATETIME, and relationship hints. The skill triggers when users ask about schema, ERDs, foreign keys, or how entities relate. Developers reach for schema-exploration when onboarding to an unfamiliar Postgres, MySQL, or SQLite database and need the agent to inspect structure instead of guessing table names. The workflow is intentionally non-destructive—list and describe only—so agents build accurate SQL and API layers on verified metadata.1.1kinstalls9Data VisualizationDeepAgents data-visualization skill helps agents pick chart types, transform tabular data, and produce frontend-ready plots and dashboards with labeling, accessibility, and export patterns for SaaS analytics surfaces.251installs10Cuml Machine LearningEnables deep agents to apply NVIDIA cuML for GPU machine-learning training and inference inside backend toolchains. It documents patterns for classification, regression, and clustering at scale when agents need model steps beyond basic dataframe analytics.231installs11RememberLangChain DeepAgents remember skill adds persistent memory to agents so conversations, facts, and task context survive across sessions, enabling longer workflows and less redundant rediscovery.213installs12Gpu Document ProcessingLangChain DeepAgents skill for GPU-accelerated document processing: ingest, parse, and prepare large document sets efficiently for agent backends, embeddings, and retrieval-heavy production workflows.210installs13Cudf AnalyticsTeaches deep agents to use NVIDIA cuDF for GPU-accelerated dataframe analytics on large tabular data. It covers backend patterns for fast filtering, aggregation, and feature preparation inside agent tooling when CUDA hardware is available.203installs14PlanningDeepAgents planning skill defines how agents break objectives into staged tasks, select tools, track dependencies, and recover from failures—core orchestration for building dependable autonomous coding and research workflows.195installs15Code ReviewDeepAgents code-review skill orchestrates automated pull-request inspection: analyze diffs, surface bugs and regressions, check conventions, and emit reviewer-style feedback aligned with team quality bars before shipping.194installs16Competitor AnalysisLangChain DeepAgents skill for systematic competitor analysis: map rivals, compare capabilities and positioning, and surface differentiation opportunities before validation or build work begins.178installs17Coding Prefscoding-prefs from langchain-ai/deepagents teaches Claude how to set and apply coding preferences for Deep Agents, aligning autonomous edits with your stack, lint rules, and team conventions.171installs