---
schema_version: '1.0'
id: news-20260715-c501bc
url: https://gotosocial.chinng-lab-srv.dev/anchor-automatic-cli-agent-hardening-through-offense-based-auditing-and-remediation/
url_hash: c501bc456a7157b8faa616366799b7da5ac1290e444c1ac522999d56adf0ee19
canonical_url: https://gotosocial.chinng-lab-srv.dev/anchor-automatic-cli-agent-hardening-through-offense-based-auditing-and-remediation
source: ghost-chinng-lab
category: news/science
category_raw: 科学・研究
region: JP
tags:
- 科学・研究
- CLI
- 監査アプローチ
- Auditing
- 自動セキュリティ
- Agent
lang: ja
published_at: '2026-07-15T00:30:50Z'
fetched_at: '2026-07-15T03:38:51Z'
updated_at: '2026-07-15T03:39:17Z'
status: published
content_hash: null
license_note: full
summary: ANCHOR は CLI エージェントのセキュリティを攻撃者視点で監査し、脆弱性を自動修復するフレームワークである。Offense-based Auditing
  で権限昇格・コマンド連携の悪用経路を検出し、Remediation でサンドボックス検証を経た安全な自動修復を実行する。ナレッジグラフと LightRAG を活用して複合的なリスク評価を行い、CLI
  環境の信頼性向上と運用負荷軽減を実現する研究。
summary_source: llm
summary_en: ANCHOR is a framework that automatically repairs vulnerabilities by auditing
  the security of the CLI agent from an attacker perspective. Off -based Auditing
  detects abusive path of privilege escalation and command collaboration, and performs
  a safe auto-repair after sandbox verification with Re。ation. A study that utilizes
  knowledge graphs and LightRAG to evaluate complex risks to improve the reliability
  of the CLI environment and reduce operational load。
entities:
- name: CLI エージェント
  type: artifact
- name: Offense-based Auditing
  type: method
- name: Agent A
  type: person
- name: ANCHOR
  type: framework
- name: Plant Automaticity Control
  type: UNKNOWN
- name: Quantum breakthroughs
  type: event
key_facts: []
related: []
related_auto:
- name: Agent B
  type: person
  weight: 1.0
- name: Agent C
  type: person
  weight: 1.0
- name: Sandbox
  type: method
  weight: 1.0
- name: RBAC
  type: concept
  weight: 1.0
- name: Knowledge Graph
  type: artifact
  weight: 1.0
title: 'ANCHOR: Automatic CLI Agent Hardening through Offense-based Auditing and Remediation'
---

# ANCHOR: Automatic CLI Agent Hardening through Offense-based Auditing and Remediation

## TL;DR
ANCHOR は CLI エージェントのセキュリティを攻撃者視点で監査し、脆弱性を自動修復するフレームワークである。Offense-based Auditing で権限昇格・コマンド連携の悪用経路を検出し、Remediation でサンドボックス検証を経た安全な自動修復を実行する。ナレッジグラフと LightRAG を活用して複合的なリスク評価を行い、CLI 環境の信頼性向上と運用負荷軽減を実現する研究。

## Key Points
- 科学・研究 / CLI / 監査アプローチ / Auditing / 自動セキュリティ / Agent

## Details
(本文なし。リンク先参照)

## Source
元記事: [ANCHOR: Automatic CLI Agent Hardening through Offense-based Auditing and Remediation](https://gotosocial.chinng-lab-srv.dev/anchor-automatic-cli-agent-hardening-through-offense-based-auditing-and-remediation/) — published 2026-07-15T00:30:50Z