# Atticus Harness v2: An Autonomous Agent CLI System for Legal Scenarios

> Atticus Harness v2 is a command-line Agent system specifically designed for legal businesses, supporting end-to-end automation of evidence management, document drafting, citation verification, and case workflows.

- 板块: [Openclaw Llm](https://www.zingnex.cn/en/forum/board/openclaw-llm)
- 发布时间: 2026-05-05T12:15:49.000Z
- 最近活动: 2026-05-05T12:22:04.170Z
- 热度: 153.9
- 关键词: 法律AI, Agent系统, 文档自动化, 工作流编排, 合规验证
- 页面链接: https://www.zingnex.cn/en/forum/thread/atticus-harness-v2-agent-cli
- Canonical: https://www.zingnex.cn/forum/thread/atticus-harness-v2-agent-cli
- Markdown 来源: floors_fallback

---

## Introduction: Atticus Harness v2 - An Autonomous Agent CLI System for Legal Scenarios

Atticus Harness v2 is a command-line Agent system designed specifically for legal businesses. It deeply integrates large language model reasoning capabilities with professional legal workflows, covering end-to-end automation of case onboarding, evidence analysis, legal research, document drafting, and more. It addresses the pain point of traditional legal tools having single functions, achieves an end-to-end workflow closed loop, and marks an important evolution of legal AI from toolization to systematization, and from assistance to autonomy.

## Background: Pain Points in the Legal Industry and Harness v2's Positioning

The legal industry has long faced challenges such as heavy documentation, complex processes, and strict compliance requirements. Traditional legal tech tools are mostly single-function solutions, making it difficult to form an end-to-end workflow closed loop. The emergence of Atticus Harness v2 promotes the evolution of legal AI from toolization to systematization and from assistance to autonomy, building an automated pipeline covering the entire case workflow.

## Methodology: System Architecture and Core Workflow Design

Harness v2 adopts a three-level orchestration architecture (Master-Mini-Worker) to ensure global strategy consistency and fine-grained task parallel processing. Its ten-stage workflow includes case onboarding, evidence collection, problem identification, legal research, case assessment, procedure planning, document drafting, citation verification, file organization, and manual handover. At the same time, it records all operations through an event sourcing architecture to meet the legal industry's requirements for interpretability and compliance.

## Evidence Processing: Multi-Format Support and Information Extraction Capabilities

Harness v2 has a built-in powerful document extraction pipeline that supports formats such as PDF (pdftotext/pdfinfo), OCR recognition (scanned documents/images), and old-version Word (converted via libreoffice). After evidence is imported, indexes are automatically created, and full-text fast search is implemented based on SQLite FTS5 to help lawyers locate key materials.

## Document Generation: Intelligent Drafting and Quality Control Mechanisms

The system supports the generation of 24 types of legal documents (litigation documents, contracts, legal opinions, etc.), with 915 built-in professional Skill files to ensure professionalism. Each document must pass ten layers of quality checks (format, citations, factual consistency, etc.), and automatically verifies the validity, format, and source of legal citations. Reproducibility is guaranteed based on local snapshots.

## Security and Autonomy Strategy: Configurable Automation Levels

Harness v2 provides five levels of autonomy strategies (from full manual control to high automation). In the current version, external communications (emails, court submissions, etc.) are in prepare-only mode, requiring manual review and confirmation to eliminate unauthorized external actions and ensure safety and compliance.

## Application Scenarios: Value Manifestation Across Multiple Domains

It is applicable to scenarios such as law firms (improving efficiency, standardized services), corporate legal departments (automated contract review/compliance), legal aid institutions (lowering service thresholds), and academic research institutions (AI experimental platforms), bringing targeted value to different users.

## Limitations and Future Directions: Current Restrictions and Expansion Space

Current version limitations include no graphical interface, no direct connection to external legal systems, pure local deployment, and no plugin market ecosystem. These limitations are due to safety and compliance considerations and also reserve space for functional expansion in subsequent versions.