DSH InsightsDeepSeek Harness 全景观察站
周报洞察

生态洞察 · 2026-W37

重点关注 · 3 个异常信号
最大账号 uckkk 独占 1286 个插件(占全生态 11%)、均分 61——模板批量号拉高规模数字,解读增长时需剔除水分
10 个插件仓库版本领先 npm 发布版本——安装侧拿到的是旧版
dsh 官方本周发布 4 个版本,其中 3 个含 breaking 变更(dsh-v0.1.5-rc.1, dsh-v0.1.5-alpha.2, dsh-v0.1.5-alpha.1)——插件作者需评估适配

DSH 生态洞察 · 2026-W37(2026/09/07~2026/09/13)

由 DSH Insights 管线 + AI 生成 · 数据快照 2026-09-15 · 启发式评估,非安全审计

1. 核心结论

  • 头部高星未发布 npm,安装侧存在真实断层:zhu1090093659/dsh-web(7564 stars)与 yjh051108/dsh-routing-suite(7183 stars)published=false,而 dsh-market 周下载 86927,说明高星与可安装分发脱节。
  • 低质高用风险集中在记忆类:doublehappy123/dsh-memory(score 57, weekly 538)与 yangyongzhen/dsh-memory(score 64, weekly 538)同名竞品同时被高频采用,用户缺乏文档与工程保障。
  • 批量账号拉高规模水分:uckkk 独占 1286 个插件、均分 61,占全生态 11%;剔除后真实规模与质量中位数都将变化。
  • 官方 breaking 发布形成适配压力:dsh 本周 4 个版本中 3 个含 breaking 变更(rc.1、alpha.2、alpha.1),插件作者须立即评估兼容性。
  • 新队列平均分下行:cohortTrend 显示 2026-09-14 创建插件平均分 72.8,低于 08-31 的 77.9 和 09-07 的 76.7,提示新进入者质量下滑。

2. 头部插件深潜

  • dsh-market/dsh-market:官方市场插件,S 级 score 95、stars 3899、weekly 86927;它构成了站内分发闭环,但同目录下 KokuYu-sysu/dsh-market-desktop(stars 1)也被同一批用户下载,品牌山寨风险高。
  • zhu1090093659/dsh-web:周下载为 null 但 stars 7564,且未发布 npm;高星未发布意味着用户无法通过官方途径安装,作者可能依赖私有分发或尚未完成发布标准。
  • yjh051108/dsh-routing-suite:stars 7183、grade B、score 88、published=false;作者 yjh051108 有 3 个插件,头部热度未转化为可发布的工程化产物。
  • omdsh-dev/DSH-better-sidebar:stars 0、score 100、weekly 73829;典型的“零星高用”,证明官方推荐/收录比社区 star 更能驱动真实使用。
  • MeteorNOX/DeepSeek-Balance-Whale-Widget:stars 2357、S 级、weekly 6405、published=true;它是少数同时具备高星、高下载、正常发布的健康头部,可作为其他作者的样板。

3. 作者格局

  • 作者总数 6384,多插件作者 1521,top10Share 15.5%,头部 star 集中度适中,但 topByStars 里 zhu1090093659 单人贡献 7564 stars,生态受个别作者影响大。
  • 高星作者:zhu1090093659(1 插件 7564)、yjh051108(3 插件 7346)、liustack(3 插件 4454);多产作者:xby-skill(175 插件、0 stars、avg 67.2)、omdsh-dev(67 插件、1816 stars、avg 79.3)。
  • 高质量作者如 SiriLee(3 插件 avg 96.7、stars 73)与 EasyTZ(4 插件 avg 95.5、stars 0)未获应有曝光,连接者 Jstn-1g(5 插件 stars 4736)也仅局部可见。
  • 信号显示 uckkk 批量号 1286 个插件、avg 61,严重污染作者榜与规模统计;整体判断:生态以流量驱动为主,社区协作网络尚未成型

4. 质量 × 采用度交叉发现

  • 被低估宝藏:omdsh-dev/DSH-better-sidebar(score 100, weekly 73829, stars 0)——新一代侧边栏底座,应尽快收录进官方 awesome 或市场精选。
  • 被低估宝藏:Ultronen/dsh-archived-chats(score 100, stars 22, weekly 586)——本地优先会话生命周期管理,方向稀缺值得关注。
  • 低质高用:doublehappy123/dsh-memory(score 57, weekly 538)与 yangyongzhen/dsh-memory(score 64, weekly 538)——同名记忆类插件评分不及格却被高频使用,用户面临数据丢失/兼容风险。
  • 低质高用:charleswin7/dsh-skill-browser(score 64, weekly 513)——文件浏览类但分数偏低,高频使用暴露在维护不足的隐患中。
  • 判断:对于作者,低星高质说明分发通道重要;对于用户,必须过滤低分插件;对于官方,需要质量闸门。

5. 生态主题地图

  • 会话与记忆主题(tags session 95、memory 68、skills 61)对应“会话管理” 923 个插件 avg 78.8,代表插件如 dsh-context(S 级 score 95)和 dsh-archived-chats。
  • 搜索与导航主题(search 85、ui 73、webui 51)对应“搜索/命令面板” 615 个 avg 78.3,dsh-market 是核心,但 KokuYu-sysu 山寨版也在此类。
  • 模型/代理主题(模型/代理 1070 个 avg 78.6)是数量第二大类,但 topPlugins 里 Minglink/dsh-infinite-gen-4(stars 1494, grade B)显示安全/红队类插件有热度,需关注合规。
  • 拥挤方向:类别“其它”有 4145 个插件但 avgScore 仅 72.1,为全类别最低,表明大量未分类、低工程标准的模板插件堆积。
  • 机会方向:每个官方场景(session-archive、file-explore、terminal-ssh 等)均只有 6 个插件,说明典型工作流深度不足,垂直领域(如终端/远程移动)存在空白。

6. 生态健康度评估

维度本期数值解读
规模11191(含 uckkk 1286 模板号)绝对规模看似过万,但约 11% 来自单一低质批量号,真实健康规模更低。
增长weeklyDiff 基线 2026-09-08 累计新增 10018 月中旬周新增曾达 4004,本周仅 728/90,增长动能明显回落。
质量分布S 272 / A 859 / B 5378 / C 3408 / D 1274B、C 级合计 78.6%,质量中位;D 级占 11.4% 不容忽视。
活跃度(7 日/30 日)active7 2773 (24.8%) / active30 11191 (100%)30 日全活跃口径可能存在数据快照偏差,但 7 日活跃仅四分之一,长尾激活不足。
npm 发布率41.8%(published 4675 / unpublished 6516)近六成插件未发布 npm,安装侧与仓库侧脱节,用户难以获取最新功能。
文档覆盖readme 11162 / zh 5333 / none 29;zhPct 47.6%readme 几乎全覆盖,但中文文档覆盖率仅 47.6%,对中文生态用户仍有障碍。

cohortTrend:最近 3 周新插件平均分 77.9→76.7→72.8 持续下滑,尤其 09-14 队列降至 72.8,说明新入局者以低质量模板批量号为主,生态面临“数量换质量”风险。

7. 异常与应对

dominant-owner · severity low

  • 可能原因(假设):uckkk 使用模板批量生成插件以抢占命名空间或刷量,缺乏人工维护。
  • 影响面:规模统计失真,用户通过列表浏览时被低质插件淹没,真实优质插件发现成本升高。
  • 应对建议:DSH 官方对单账号插件数设置阈值并人工审计模板批量号;作者榜与热门榜剔除该账号后重新计算。

npm-stale · severity low

  • 可能原因(假设):作者在 GitHub 上快速迭代但未同步发布 npm,或 npm 账号/CI 流程断链。
  • 影响面:用户通过 npm 安装到旧版本,可能遇到已修复的 bug 或缺失的新功能;npmStaleTop 中 Aegis 仓库 v2.9.6 vs npm 0.1.0 差异极大。
  • 应对建议:插件 author 手动或通过 CI 触发 npm publish;官方在插件详情页展示 npm 版本与仓库版本差异提醒。

breaking-release · severity high

  • 可能原因(假设):DSH 核心在 v0.1.5 系列进行 API 调整,alpha/rc 阶段引入不兼容变更,属正常迭代但发布节奏紧凑。
  • 影响面:已发布插件可能在新版 DSH 上崩溃或失效;作者需快速适配,否则用户安装后面临故障。
  • 应对建议:官方提前 2 周发布 breaking change 预告与迁移指南;插件作者立即在测试环境运行 dsh-v0.1.5-rc.2 并修复 manifest/API 调用。

8. 下一阶段重点与建议

  • 插件作者:针对 dsh-v0.1.5 系列 breaking 变更跑通兼容性测试;补齐 npm publish 与版本同步;为中文用户补充 zh 文档;放弃模板批量发布,转向细分场景开发。
  • 生态用户:优先选择 S/A 级且 published=true 的插件;对 memory 类低分高用插件保持警惕,检查其 README、测试和最近更新时间;关注官方市场精选而非仅看 star 数。
  • DSH 官方:清理 uckkk 等批量号并在规模统计中剔除;对 quadrant.risks 中高下载低分插件引入人工复核或质量提示;建立 breaking release 的提前通知与迁移文档;提高高星未发布插件(如 dsh-web、routing-suite)的发布转化率。

9. 官方动态解读

  • dsh 官方仓库 223,959 stars,本周发布 v0.1.5-rc.2(非 breaking),但 rc.1、alpha.2、alpha.1 为 breaking,插件生态进入适配窗口期。
  • DeepSeek-V3(104,448 stars)与 R1(91,991)仍是流量基本盘,awesome-deepseek-integration(39,117)表明聚合生态需求强,dsh 应加强官方 awesome 收录。
  • dsh-market 周下载 86,927 且 S 级,但桌面版 KokuYu-sysu/dsh-market-desktop 同名山寨同量级下载、stars 仅 1,官方需处理品牌混淆。
  • 官方最新 release 节奏快(4 个版本/周),无测试 CI 的插件作者将面临持续适配压力,拥有 CI 的头部插件将拉开差距。

DSH Ecosystem Insights · 2026-W37(2026/09/07~2026/09/13)

Generated by the DSH Insights pipeline + AI · snapshot 2026-09-15 · heuristic evaluation, not a security audit

1. Core Findings

  • High-star plugins remain unpublished, breaking install-side continuity: zhu1090093659/dsh-web (7564 stars) and yjh051108/dsh-routing-suite (7183 stars) have published=false, while dsh-market logs 86,927 weekly downloads.
  • Low-quality high-usage clustering in memory tools: doublehappy123/dsh-memory (score 57, weekly 538) and yangyongzhen/dsh-memory (score 64, weekly 538) are near-identical low-scoring plugins under heavy use.
  • Batch account inflates ecosystem scale: uckkk controls 1,286 plugins (11% of total) with avg score 61; excluding it materially lowers true size.
  • Breaking releases create adaptation pressure: 3 of dsh's 4 releases this week (rc.1, alpha.2, alpha.1) contain breaking changes.
  • New cohort quality trending down: cohortTrend avgScore fell from 77.9 (08-31) to 76.7 (09-07) to 72.8 (09-14), signaling entry-level quality decline.

2. Head Plugin Deep Dives

  • dsh-market/dsh-market: official market plugin, S-grade score 95, stars 3899, weekly 86927; it anchors distribution, but clone KokuYu-sysu/dsh-market-desktop (stars 1) captures identical download volume, posing brand risk.
  • zhu1090093659/dsh-web: 7564 stars but weekly null and unpublished; users cannot install via official npm, indicating distribution gap.
  • yjh051108/dsh-routing-suite: 7183 stars, grade B, unpublished; author yjh051108 has 3 plugins, but popularity hasn't translated into publish-ready engineering.
  • omdsh-dev/DSH-better-sidebar: 0 stars, S score 100, weekly 73829; a zero-star high-usage case where curation trumps community stars.
  • MeteorNOX/DeepSeek-Balance-Whale-Widget: 2357 stars, S-grade, weekly 6405, published; a healthy template combining visibility, adoption, and npm release.

3. Author Landscape

  • 6384 total authors, 1521 multi-plugin authors, top10Share 15.5%; overall moderate concentration, but zhu1090093659 alone holds 7564 stars.
  • Top-star authors: zhu1090093659 (1 plugin, 7564), yjh051108 (3, 7346), liustack (3, 4454); prolific: xby-skill (175 plugins, 0 stars, avg 67.2), omdsh-dev (67, 1816, avg 79.3).
  • Quality authors like SiriLee (3 plugins avg 96.7, stars 73) and EasyTZ (4, avg 95.5, stars 0) remain underexposed; connector Jstn-1g (5 plugins stars 4736) also limited visibility.
  • Signal shows uckkk batch account with 1286 plugins, avg 61, severely polluting author charts. Judgment: ecosystem is traffic-driven rather than community-driven, with sparse collaboration network.

4. Quality × Adoption Cross-Findings

  • Underrated gem: omdsh-dev/DSH-better-sidebar (score 100, stars 0, weekly 73829) — next-gen sidebar base, should be officially curated.
  • Underrated gem: Ultronen/dsh-archived-chats (score 100, stars 22, weekly 586) — local-first session lifecycle, scarce direction.
  • Risk: doublehappy123/dsh-memory (score 57, weekly 538) and yangyongzhen/dsh-memory (score 64) — failing scores with heavy traffic, exposing users to data/compat risks.
  • Risk: charleswin7/dsh-skill-browser (score 64, weekly 513) — low score, high usage, maintenance hazard.
  • So-what: for authors, distribution channels outweigh stars; for users, score filtering is mandatory; for officials, quality gates are overdue.

5. Ecosystem Theme Map

  • Session & memory cluster (tags session 95, memory 68) maps to category 会话管理 923 plugins avg 78.8; representatives dsh-context and dsh-archived-chats.
  • Search & nav cluster (search 85, ui 73) maps to 搜索/命令面板 615 avg 78.3; dsh-market is core but clone pollution exists.
  • Model/proxy cluster (模型/代理 1070 avg 78.6) sees red-team tools like Minglink/dsh-infinite-gen-4 gain heat; compliance needs monitoring.
  • Crowded: “其它” category holds 4145 plugins avgScore 72.1, lowest; accumulation of unclassified template plugins.
  • Opportunity: each official scenario (session-archive, terminal-ssh, etc.) has only 6 plugins, indicating shallow vertical coverage and whitespace in terminal/remote.

6. Ecosystem Health Assessment

DimensionThis Period ValueInterpretation
Scale11191 (incl. uckkk 1286 batch)Size looks >11k, but ~11% from one low-quality batch account; true healthy scale lower.
GrowthweeklyDiff baseline 2026-09-08 cumulative +1001Mid-August weekly additions hit 4004; this week 728/90, momentum declining.
Quality DistributionS 272 / A 859 / B 5378 / C 3408 / D 1274B+C account 78.6%, median quality; D at 11.4% not trivial.
Activity (7d/30d)active7 2773 (24.8%) / active30 11191 (100%)30d all-active may be snapshot artifact; 7d only quarter active suggests weak long-tail activation.
npm Publish Rate41.8% (published 4675 / unpublished 6516)Nearly 60% unpublished; install side lags repo side.
Docs Coveragereadme 11162 / zh 5333 / none 29; zhPct 47.6%README nearly universal, but Chinese docs only 47.6%, barrier for Chinese users.

cohortTrend: avgScore of recent cohorts fell 77.9→76.7→72.8 over three weeks, with 09-14 hitting 72.8, indicating an influx of low-quality template accounts and a quantity-over-quality risk.

7. Anomalies and Responses

dominant-owner · severity low

  • Possible cause (hypothesis): uckkk uses templates to mass-generate plugins for namespace squatting or inflated metrics, lacking maintenance.
  • Impact: scale metrics distorted, users flooded with low-quality plugins, discovery cost rises.
  • Recommendation: DSH official sets per-account plugin thresholds and manually audits batch accounts; recompute charts after excluding such accounts.

npm-stale · severity low

  • Possible cause (hypothesis): fast GitHub iteration without npm sync, or broken npm/CI pipeline.
  • Impact: users install outdated versions; Aegis repo 2.9.6 vs npm 0.1.0 exemplifies severe drift.
  • Recommendation: authors automate npm publish via CI; official displays repo/npm version mismatch prominently.

breaking-release · severity high

  • Possible cause (hypothesis): DSH core API changes in v0.1.5 series, normal but compressed release cycle.
  • Impact: published plugins may break on new DSH; authors need rapid adaptation.
  • Recommendation: official provides 2-week breaking change notice and migration guide; authors test against v0.1.5-rc.2 immediately.

8. Next Steps by Role

  • Plugin authors: run compatibility tests against dsh-v0.1.5 breaking changes; sync npm publish with repo; add Chinese docs; abandon template mass release, focus on niches.
  • Ecosystem users: prefer S/A+published=true plugins; scrutinize memory category low-score high-usage plugins for docs/tests/recent updates; rely on official curation, not star count alone.
  • DSH official: purge uckkk-type batch accounts from metrics; add quality review for quadrant.risks high-download low-score plugins; establish breaking release notification; improve publish conversion for high-star unpublished plugins like dsh-web and routing-suite.

9. Official Dynamics

  • dsh repo has 223,959 stars, released v0.1.5-rc.2 non-breaking, but rc.1/alpha.2/alpha.1 breaking; ecosystem enters adaptation window.
  • DeepSeek-V3 (104,448) and R1 (91,991) remain traffic base; awesome-deepseek-integration (39,117) shows strong aggregation demand; dsh should strengthen official awesome curation.
  • dsh-market weekly 86,927 S-grade, but desktop clone KokuYu-sysu/dsh-market-desktop gets same downloads with 1 star; official must resolve brand confusion.
  • Official release cadence of 4 versions/week puts pressure on plugins without CI; those with CI will pull ahead.