
Security News
Anthropic Identifies Biased Reasoning and Recklessness as Drivers of Claude’s PyPI Attack
Anthropic found biased reasoning and recklessness drove Claude Mythos 5 to publish malware on PyPI and compromise a security vendor.
dsh-decision-split
Advanced tools
决策分屏:agent 输出含「决策摘要」段的方案时,主窗口显示摘要与决策编辑框,右侧详情栏独立展示完整 markdown,决策一键发回 agent。
决策分屏插件 —— DeepSeek Harness Web UI 的决策工作台。
当 agent 输出包含「决策摘要」段(任意标题层级,如 ## 决策摘要、### 决策要点)的方案时:
【决策】… 前缀作为一条消息回传给 agent,让它基于你的决策继续推进。视觉完全对齐 DSH Web UI 原生设计变量(--dsw-* tokens、MarkdownText 字体规范、原生按钮样式)。
从 npm 安装(推荐):
dsh plugin --profile web add dsh-decision-split
或从 GitHub release 安装:
dsh plugin --profile web add https://github.com/yuanbaoerer/dsh-decision-split/releases/latest/download/dsh-decision-split-0.1.2.tgz
安装后重启 dsh web 生效。
让 agent 在输出方案时,用固定格式列出待决策点,例如在指令里加一句:
输出方案时,最后用
## 决策摘要段列出所有需要我拍板的决策点。
插件识别的标记(任意标题层级 ##–######、忽略大小写、允许结尾冒号):
决策摘要 · 决策要点 · 待决策清单 · 决策清单 · 决策事项 · Decision Summary · Decision Points
dsh-decision-split/
package.json # dsh.bundle.patch + dsh.client 清单(web 客户端插件)
cordis.patch.yml # bundle patch:insert 自身
lib/index.js # host 半(空壳)
lib/client.js # client 半(window.__ModuleLoader__ bundle)
修改 lib/client.js 后重启 dsh web 生效。发布新版本(先 bump package.json 的 version 并打 tag):
npm pack # 生成 dsh-decision-split-<version>.tgz
gh release create v<version> dsh-decision-split-<version>.tgz # GitHub release
npm publish # npm registry(需带 bypass-2FA 的发布 token)
MIT
FAQs
决策分屏:agent 消息尾部提供「在侧栏查看全文」入口,右侧详情栏独立滚动展示该条消息的完整 markdown,边看全文边回主窗口操作。
We found that dsh-decision-split demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 1 open source maintainer collaborating on the project.

Security News
Anthropic found biased reasoning and recklessness drove Claude Mythos 5 to publish malware on PyPI and compromise a security vendor.

Research
/Security News
Malicious Chrome and Firefox extensions target Axiom Trade and Padre users, stealing session tokens and wallet data.

Security News
GPT-6 Astra hits 100% on ExploitBench and finds zero-days autonomously, while independent tests reveal scope violations and monitoring gaps.