loeanxi/dsh-injection-guard1

dsh-injection-guard

Source-aware prompt injection protection for DeepSeek Harness

AI 분석

核心用途是防御提示词注入攻击,对敏感调用进行安全审查与拦截。适合对安全性要求极高、需要防止模型被恶意指令劫持的生产环境或自动化任务。

패키지
dsh-injection-guard
버전
0.1.0
최근 업데이트
2026. 8. 21.

설치

$npx -p @deepseek-ai/dsh dsh plugin --profile web add github:loeanxi/dsh-injection-guard

Configuration

All options are optional. The safe defaults are shown below:

OptionDefaultPurpose
logtrueEmit an explainable audit message for each decision
askThreshold60Minimum score for ASK when a block threshold is not reached
failClosedtrueReview-gate sensitive calls when no turn state is available
semantictrueAdd conservative local intent signals; never weakens deterministic rules
localeenUse en or zh-CN for audit text while retaining machine markers

For Chinese audit output, add locale: zh-CN to the plugin configuration:

config:
  log: true
  locale: zh-CN

配置项

所有配置项均可省略,以下是安全默认值:

配置项默认值作用
logtrue为每次决策输出可解释审计信息
askThreshold60未达到阻断阈值时触发 ASK 的最低分数
failClosedtrue没有 Turn 状态时,对敏感调用进行人工复核
semantictrue增加保守的本地意图信号,不会削弱确定性规则
localeen使用 enzh-CN 输出审计文本,同时保留机器标记

如果希望页面上看到中文审计结果:

config:
  log: true
  locale: zh-CN