Red-team engagement mode: send one target organization name; a planner session runs five execution roles (recon, asset triage, vulnerability discovery, exploitation, internal pivot) with configurable concurrency; preflight checks skills/resources; scoring uses four stages (recon, internet, boundary, internal) with server-side caps and privilege-wins dedup; findings stay in a local SQLite fact base. The persistent right-side console has nine pages and includes asset mapping, agents, sessions/tunnels, findings, score targets, reports, POC knowledge base and skill library. v0.15.2 adds automatic HTTP evidence/action records, batch ingest, scope check, tunnel pivot probe, data proof, cleanup ledger and T0-T3 asset admission to prevent scope drift. 59 redteam_* tools; 38 bundled skills.
Install
# from npm (prebuilt)
dsh plugin --profile web add dsh-redteam-mode
# from GitHub (first run asks for allowBuilds approval — follow the hint, retry)
dsh plugin --profile web add github:Jueze-2019/dsh-redteam-mode#path:/packages/redteam-bundle
Any plugin you install runs third-party code with your own permissions — it can read your files, use your credentials, and reach the network, and tool approvals don’t sandbox it. GitHub-sourced plugins also run build scripts at install time — pnpm blocks those until you allow them, so an install can stop with ERR_PNPM_GIT_DEP_PREPARE_NOT_ALLOWED or ERR_PNPM_IGNORED_BUILDS; dsh prints the exact key to add under allowBuilds in your profile’s pnpm-workspace.yaml, and the install works on the next run. Allowing a build is a trust decision: only install sources you trust, and pin a commit (github:owner/repo#sha).
README
This plugin publishes its README in Chinese only.
DeepSeek Harness 的红队作战模式插件:给一个靶标单位名称,就拉起一支红队智能体 (信息收集 → 漏洞检测 → 漏洞利用 → 内网渗透),所有发现落进本机 SQLite 事实库, 右侧常驻控制台实时看资产、得分与报告。
安装
dsh plugin --profile web add dsh-redteam-mode
装完重启一次 dsh web,然后在设置 → 插件市场里就能看到它(也可以直接在上面搜
redteam)。插件自带 红队模式 与 39 个技能(含 AD/云/网络设备/数据/C2 等专项技能),不需要手动拷任何文件。
版本对应:DSH 0.2.1 及以上用 0.15.6 及更新的版本;DSH 0.1.7–0.2.0 用 0.12.1–0.13.1;DSH 0.1.6 及更低用 0.12.0 及更早版本。0.12.0 及更早的插件在 DSH 0.1.7+ 上,新建会话的模式列表里不会出现「红队模式」(资产库、控制台、工具与技能都不受影响)。
DSH 0.2.1 的破坏性变更:
@deepseek-ai/dsh-tool-subagent移除了toolFilter/persona, 再写会在apply()直接抛错toolFilter is no longer supported,subagent委派工具不再注册。 0.13.2 起红队预设改用独立的redteam-worker执行角色预设 +agentPreset;升级 DSH 后请同时 升级本插件。旧版 DSH 仍可加载本版(未知字段被忽略),但子智能体的叶子约束只剩maxDepth: 1。
使用:新建会话 → 选预设 红队模式 → 直接发靶标单位名称。
关于工具链:本包包含全部 39 个技能;不包含任何自动下载/安装安全工具的脚本、二进制、安装链接或教程。工具由用户自行准备。
从预发布期的「开发版」升级过来
如果你在 0.7.0 之前用过开发版(在 ~/.dsh/profiles/web/cordis.patch.yml 里手写过
redteam-store / redteam-ui 两行,指向 dsh-redteam-store / dsh-redteam-ui),
装完本插件后启动会报服务/条目冲突,先跑一次迁移把旧行清掉:
node "$(npm root -g 2>/dev/null)/dsh-redteam-mode/lib/migrate-legacy-rows.mjs" --dry-run # 预演
node node_modules/dsh-redteam-mode/lib/migrate-legacy-rows.mjs # 在 profile 目录里执行
它会只删那两条(连子行),其它内容与注释原样保留,写回前自动备份;已经在用 0.7.0+ 或者新装的机器不需要跑。
它给你什么
| 能力 | 说明 |
|---|---|
| 六个执行角色 | 信息收集 / 资产梳理 / 漏洞发现 / 漏洞利用 / 内网信息收集 / 内网渗透,主会话只做调度与指挥 |
| 角色工具裁剪 | 6 个 redteam-worker-* 预设按 profile 裁剪工具(recon 38 / assess 40 / vuln-scan 47 / exploit 53 / internal-recon 38 / internal 54;默认全部 59) |
| 资产测绘 | C 段按内外网分组、端口/服务/指纹/域名/Web 标题、易打性评估、测试状态 |
| 得分目标 | 得分点可编辑、命中记录一行一条;自建账号不计分;账号权限与数据库权限按「同一资产 + 同一端口」封顶(拿到最高权限账号即该服务拿满,同服务再刷账号标「服务已拿满 · 不计分」),其余得分点仍不设数量上限 |
| 会话隧道 | WebShell 与隧道状态、连通性实测、是否跨越靶标边界的判定 |
| 报告 | 按攻击链顺序、按阶段可折叠,每条得分附可直接粘进 Yakit 的原始请求;自建账号与同服务重复命中整条剔除(抬头给出剔除条数) |
| POC 知识库 | 全局共享(跨靶标):先查库再联网/手搓,验证有效的回填;同时检索本机 nuclei 模板库 |
| 技能 | 39 个技能随 npm 包分发(含 AD/云/网络设备/数据/C2 等专项技能);不打包任何工具二进制、安装脚本或安装链接/教程 |
数据放在哪
- 全部落在本机
$DSH_HOME/redteam/:engagements/<靶标>/(SQLite 事实库 +runs/证据 + 按目标分目录的攻击文件)、knowledge.db+pocs/(跨靶标共享的 POC/EXP 知识库)。 - 不含任何遥测、上报或云端同步;凭据明文只在本机库里,面板直显便于复用。
工具(59 个 redteam_*)
资产/端口/服务/指纹/漏洞/凭据/会话/得分/报告/知识库全覆盖。典型入口:
redteam_engagement_open 绑定本次演练靶标
redteam_asset_add 落库一个资产(端口带 service/product/version/url/title)
redteam_batch 批量资产/评估/漏洞入库(assets≤500 / assess·vulns≤200),减少逐条调用
redteam_scope_check 核对 IP/URL 是否在靶标授权 scope 内
redteam_vuln_add 落库一个漏洞(带 severity/status/gained)
redteam_score_hit 记一次得分(command/result 自动生成动作记录;自建账号不计分;同服务只算一次)
redteam_data_proof P1 敏感数据结构化证明(字段/条数/样本/来源,≥1 万条标 meets_threshold)
redteam_score_report 生成「四段式红队得分报告」
redteam_sessions 一屏看 WebShell / 隧道 / 凭据 / 会话
redteam_pivot_probe 经 SOCKS5 实测多个内网 host:port 的可达矩阵
redteam_cleanup 上传的马/临时账号/隧道/文件清理台账
redteam_poc_search 打 Nday/1day 前先查知识库 + 本机 nuclei 模板库
redteam_poc_add 把验证有效的通用 POC/EXP 回填知识库(脱敏)
从源码构建
本包是自包含的:lib/ 由 tools/build.mjs 从同仓库的三个源码包生成
(redteam-store / redteam-tools / redteam-ui),预设与技能也一并拷进来。
node packages/redteam-bundle/tools/build.mjs # 生成
node packages/redteam-bundle/tools/build.mjs --check # 只校验是否漂移(CI 用)
node packages/redteam-bundle/test/bundle.test.mjs # 打包契约 + 泄露 + 自举自检
授权与合规
仅在获得明确书面授权的范围内使用。本插件是作战工具:它本身不替你做授权判断, 请自行确认目标范围、时间窗与允许手段,并遵守所在司法辖区法律。
Links
More in this category
kenryu42/cc-safety-net★ 1749
Pre-execution guard that blocks destructive Git and filesystem commands and access to secrets such as SSH keys and .env files before a tool call runs.
toby-bridges/api-relay-audit★ 887
Runs local security audits of AI API relays and LLM proxies from DeepSeek Harness, producing Markdown reports for prompt injection, model substitution signals, tool-call rewriting, error leakage, stream integrity, and profile-gated Web3 risks.
SeaOf0/dsh-redteam-model★ 741
Authorized-security DSH collection: nine work modes (redteam coordinator, pentest, code audit, binary analysis, attack-defense, AV evasion, incident response, cloud security, CTF solving) and fifteen runtime plugins, managed from a settings page with one-click deploy, install, update and uninstall.
howmp/dsh-pentest★ 608
Authorized pentest mode for DeepSeek Harness — exploration chain, assets and findings with a Web view.
ADWMC/helm-d#helmd★ 443
Security-analysis plugin for DeepSeek Harness: 33 tools spanning six security domains, evidence/reporting and toolbox, plus the case lifecycle, routing and tool ledger; a session-bound case workflow with an evidence chain, E-id validation and reproducible findings; 362 on-demand references; three hot-swappable persona profiles; and a four-tab workbench (armor / case / intercept / about) with case-state recovery after context compaction.
PerryLink/dsh-auto-review★ 234
Second-model auto-review on the approval answerer chain: a read-only reviewer subagent returns structured allow/deny verdicts with reasons, fail-closed by default.
Community comments
Comments are public GitHub Discussions. Loading them connects to GitHub and Giscus; a GitHub account is required to post.