dsh-auto-review：DeepSeek Harness 的自动审批判断插件
`dsh-auto-review` 是 DeepSeek Harness 的 MIT 许可插件，专为解决 `approval policy` 设为 `'auto'` 时超出 `workspace-write` 的沙箱升级审批问题。它作为 answerer 接入 `approval/request`，采用“确定性过滤 + LLM 审查”的双重机制实现自动决策，无需人工介入。 核心能力包括：首先执行确定性过滤，通过 `hardline` 和正则规则即时拒绝或允许明确命令；灰区请求则进入 clean-context LLM 安全审查，仅基于工具名、理由和命令进行判断，避免携带完整会话上下文。插件支持
Read MoreDSH 插件 dsh-auto-approval-plugin：在 Workspace Write 与 Full access 之间加入自动审批档
StyxNether 开发的 dsh-auto-approval-plugin 是 DeepSeek Harness (DSH) 的权限审批插件，旨在填补 Workspace Write 与 Full access 之间的权限档位空缺。该插件引入 auto-approval 预设，在审批路径上注册监听器，基于 session log 中的真实工具参数（而非模型生成的 justification）进行自动判定。若请求目标位于用户配置的可信区（支持 realpath 解析以防符号链接绕过）或匹配无害模式，插件将自动批准该次调用（allowed-once）；否则，请求将传递给人工确认，插件本身不执行
Read MoreDSH-Approval-Gate: DeepSeek Harness Automatic Approval Gating Plugin
dsh-approval-gate is an MIT-licensed plugin for DeepSeek Harness, designed to balance the efficiency of automated approval with security. It leverages Flash models to perform risk pre-assessment on sandbox escape operations, categorizing them into SAFE and RISKY types. For reversible, low-risk operations, the plugin uses a confirmation-based learning mechanism to automatically approve them, reducing repetitive manual confirmations. For the five categories of hard-risk operations—deletions, credential handling, remote/production environments, system paths, and irreversible batch operations—it always routes to manual confirmation to ensure critical security. The plugin provides a manual review UI to display the approval timeline and supports hot-reloading of configurations. Version v0.5.0 introduces file change comparison and one-click revert functionality, allowing users to view unified diffs and restore files based on session snapshots. Users need to install it via the command line.
Read Moredsh-auto-approve: Adding Auto Permission Level Between Workspace Write and Full Access
dsh-auto-approve inserts an auto permission tier between workspace-write and danger-full-access for DeepSeek Harness. The sandbox boundary is identical to workspace-write, with routine sandbox escalations approved in one go by a classification model; cases hitting a deterministic danger list, where the classifier is uncertain, or due to timeout or malformed responses are still escalated to manual approval. The plugin supports Web, TUI, and Desktop frontends and can be installed via dsh plugin and enabled by activating the Auto tier in Permissions.
Read More