Use a strict process to upgrade AI from a “typist” to an “engineer/architect,” emphasizing TDD, division of responsibilities, and acceptance criteria.

TDDWorktreeReviewDiscipline

agency-agents

Organized anthropomorphic enterprise team

Bring the "organization" into the repository: a large number of role templates with metrics and deliverables, collaborating by department.

OrgMetricsDeliverablesPlaybooks

get-shit-done

GSD anti-context-corruption system

Engineering countermeasures against Context Rot: planning → wave-based isolated execution → acceptance → atomic commit.

ContextPlanExecuteVerify

TradingAgents

Macroeconomic environment financial trading agent

Multi-agent collaboration + debate/reflection + risk-control closed loop to make more robust trading decisions.

Multi-AgentDebateReflectionRisk

skills

Minimalist Entrepreneur skill library

10 startup companion skills based on The Minimalist Entrepreneur: prioritize validation and oppose building before selling.

EntrepreneurValidationPlaybookDecision

follow-builders

Information refining and retrieval agent

Track top builder information sources, distill them into a noise-reduced, readable digest, and support multilingual typesetting and delivery.

DigestSummarizeTranslateDelivery

skill-from-masters

Master-level skill extraction forge

Emphasize Skill generation with “case-backed” support: first compare real-world success and failure cases, then summarize patterns and derive actionable Skills.

CasesInductionTaxonomySkill

Framework / SDK (orchestration layer)

First learn “how to build with blocks,” then learn “how to make systems stable”

LangChain

Framework / SDK (orchestration layer)

An orchestration framework for building LLM applications and Agents: chained workflows, tool calling, RAG, and multi-component integration.

FrameworkOrchestrationAgentsRAG

RAG framework / engine / platform

Split by module: data and indexing / retrieval and reranking / orchestration / engine-ification / platform delivery

Data and indexing (Ingestion / Parsing / Indexing)

Convert raw data into a searchable index structure

LlamaIndex

RAG / Document Agent framework

A RAG architecture focused on the “data and index layer”: connecting data sources, building indexes, retrieval, and document-oriented agent workflows.

RAGIndexingConnectorsAgents

Retrieval and reranking (Retrieval / Rerank)

Retrieval, fusion, and filtering to make context “more accurate and more stable”

R2R

Production-grade retrieval system / Agentic RAG

Build RAG from the perspective of a "retrieval system": provide API/SDK and more engineering-oriented retrieval and context management capabilities.

RetrievalAPIAgentic RAGSDK

txtai

Semantic search + RAG workflow

Integrated semantic search framework: combines vector retrieval, pipelines, and LLM workflows to quickly build RAG applications.

Semantic SearchRAGEmbeddingsWorkflows

Orchestration and pipelines (Orchestration / Pipelines)

Use a controllable process to connect retrieval, routing, and generation into a system

Haystack

RAG / Pipeline orchestration framework

Modular Pipeline/Agent framework: retrieval, routing, memory, and generation are explicitly controllable, making it suitable for production-grade RAG.

RAGPipelinesRoutingProduction

Engine-based implementation (RAG Engine)

An end-to-end RAG engine that feels closer to a product

RAGFlow

Open-source RAG engine (with Agent capabilities)

A practical RAG Engine: emphasizes the quality of the context layer, the combination of retrieval and answer generation, and provides a more complete productized form.

EngineRAGContextAgentic

Platform delivery (Platform / Ops)

An application-layer platform for team use and deployment operations

Dify

RAG application/workflow platform (more product-oriented)

Application-layer platform: knowledge base + workflow/Agent orchestration + operations deployment, suitable for quickly turning RAG into a usable product.

PlatformKnowledge BaseWorkflowOps

Classic agent prototype

Understand early municipal paradigms and historical evolution

AutoGPT

Classic Agent prototype (early autonomous entities)

An early representative implementation of an "autonomous Agent": a pioneering project for task loops, tool use, and multi-step execution paradigms.

AutonomousTask LoopPrototypeTools

Want to turn a popular project into your own capability?

We recommend choosing a project in the "Practical" section to follow along with, or using the "Super Individual" section to productize your skills.

Move on to the practice section Go to the super-individual section