生态洞察 · 2026-W42

重点关注 · 5 个异常信号

中账号 uckkk 有 1286 个插件、均分仅 56——疑似批量刷库/模板复制

中账号 CAI-MH 有 9 个插件、均分仅 57、1 个无 README——疑似批量刷库/模板复制

低最大账号 uckkk 独占 1286 个插件（占全生态 8%）、均分 56——模板批量号拉高规模数字，解读增长时需剔除水分

低10 个插件仓库版本领先 npm 发布版本——安装侧拿到的是旧版

高dsh 官方本周发布 5 个版本，其中 2 个含 breaking 变更（dsh-v0.2.1-rc.1, dsh-v0.2.1-alpha.3）——插件作者需评估适配

DSH 生态洞察 · 2026-W42（2026/10/12～2026/10/18）

由 DSH Insights 管线 + AI 生成 · 数据快照 2026-10-19 · 启发式评估，非安全审计

核心结论

生态规模虚火与有效参与度背离：15681 个权威插件中仅 17.6% 近 7 日活跃、44.2% 近 30 日活跃，而 uckkk 一个账号独占 1286 个均分 56 的插件，规模数字需去水后判断。
质量中位勉强健康但尾部风险不小：avgScore 73.7，S/A 级合计仅 648 个（4.1%），C/D 级占 48.7%；topDownloads 前 10 中仍有 3 个 C 级低质高用。
官方 breaking 发布与 npm 漂移形成双重适配压力：dsh 本周 5 个发布中 2 个含 breaking 变更；10 个插件仓库版本领先 npm 最新版。
高下载流量榜与高质量宝藏榜严重脱节：周下载第一的 billion-context 仅 B 级 86 分，而 quadrant.gems 中 8 个 S 级 100 分插件 stars 均 <30。
“状态/监控/用量”赛道被投机复制，合规与质量风险并存：dsh-whale-widget 及多个低质衍生品占据下载榜 5 席，YuJunZhiXue/dsh-purge 等高 star 破甲工具未发布 npm。

头部插件深潜

Ebony-Vinyl/dsh-our-free-model：免登录免 Key 调用 DeepSeek V4.1 Flash / Kimi K3 等模型，24760 stars、A 级 93 分、周下载 7921；明星光环下实际周下载不及 dsh-market 的 1/14，说明“猎奇流量”远大于持续使用。
dsh-market/dsh-market：官方插件市场入口，A 级 90 分、周下载 117724，是下载榜第三；但 npm 版本为 1.66.42 而仓库版本仅 1.43.0，存在版本漂移，用户从 npm 安装拿到的是落后 5 个版本的旧市场。
ranxianglei/billion-context：通用上下文压缩代理，B 级 86 分、周下载 450525，高居下载第一；stars 仅 1879，流量与声望倒挂，推测是刚需场景或推荐位放大，需官方确认真实来源。
MeteorNOX/DeepSeek-Balance-Whale-Widget：余额监控小鲸鱼娘，A 级 90 分、周下载 55242；它带动了至少两个低质二创（621758/dsh-whale-widget-pet-feed C 级 81 分、shixun926-dotcom/dsh-whale-widget-desktop C 级 71 分），原版优质但衍生品治理需跟上。
YuJunZhiXue/dsh-purge：模型破甲/越狱工具，7078 stars、B 级 78 分、未发布 npm；weeklyDiff 显示其一周 star 增量 +3785，虽流量暴涨但合规风险与不发布 npm 构成双刃剑。

作者格局

作者总数 9058，多插件作者 2270；头部 10 位作者 star 占比仅 2.9%，但 Ebony-Vinyl 一人持有 24862 stars、占 topByStars 前列，说明生态由少数明星插件而非作者矩阵驱动。
高质作者如 Fish-under-sea（4 插件平均 96.5 分、13 stars）、EasyTZ（4 插件平均 91.8 分、1 star）几乎无曝光；而 prolific 批量作者 xby-skill 发布 175 个插件、总 star 仅 2、均分 62.2。
uckkk 账号被 rules 检测出 1286 个插件、均分 56，占全生态 8%，直接污染作者榜与增长统计；CAI-MH 小集群同样疑似批量刷库。
连接者节点 Grivn、qikairo7、Jstn-1g 等同时拥有多个插件和较高 stars（如 qikairo7 3 插件 7881 stars），但整体 connector 网络稀疏，协作多集中在少数几人。
判断：该生态当前是流量驱动为主、社区驱动萌芽——明星单款获客，批量号制造规模，真实高质量贡献者缺流量扶持。

质量 × 采用度交叉发现

被低估的宝藏：SenmuuuuW/dsh-whale-report（S 级 100 分、周下载仅 240）能从会话事件日志生成日报/周报，是数据叙事方向的高质插件；EasyTZ/dsh-git（S 级 100 分、周下载 51）提供 Git 面板，符合开发者工作流刚需；建议官方收录并给予搜索位。
被低估的宝藏：sunzhentao/dsh--prompt--enhance（S 级 100 分、周下载 599）专注提示词增强，在“模型/代理”赛道有差异化价值；Fish-under-sea/dsh-agent-teams-fish（S 级 100 分、周下载 1663）是 AgentTeams 协作补充版，但 star 仅 4。
低质高用风险：Bald0Wang/dsh-imggenerate（C 级 62 分、周下载 6040）图像生成工具无测试/CI 与许可证，用户可能面临维护中断或安全风险；LuniteGlaze/dsh-plugin-catalog（C 级 64 分、周下载 2686）描述为空，功能不透明。
低质高用风险：ben7am1n/dsh-browser 与 duyefeng/dsh-browser（均 C 级 60-61 分、周下载 775）为同名浏览器自动化，竞争激烈且质量低，用户选型易被下载量误导。

生态主题地图

主题 1：上下文与记忆：topTags 中 session（95）、memory（68）突出，代表插件 billion-context 周下载 45 万，dsh-context（A 级 90 分）也在下载榜，上下文压缩/记忆管理是刚需主线。
主题 2：搜索与工作区增强：search（85）、ui（73）、sidebar（49）组成了另一主流，dsh-market 与 dsh-better-sidebar（S 级 95 分）是代表；官方 scenarios 中 global-search、sidebar-workspace 各仅 6 个插件，表明该方向尚未拥挤。
主题 3：模型接入与代理编排：model/agent/mcp/skills 标签合计约 250 个，MichengAI/dsh-skills-manager（B 级 88 分）与 dsh-inherit（A 级 94 分）等在做 Agent Skills 管理；但代理类插件质量方差大。
主题 4：用量与余额监控：balance（50）、usage（50）标签集中，dsh-whale-widget 及衍生品是主要代表，但大量二创拉低该主题平均质量，属于“拥挤但不出彩”的方向。
空白机会：remote-mobile、terminal-ssh、memory-knowledge 三个场景每个仅 6 个插件，而 LLM 能力标注覆盖率仅 1823/15681（11.6%），大量插件未被主题发现——官方应提升标注覆盖并鼓励这些低密度方向。

生态健康度评估

维度	本期数值	解读
规模	authoritative 15681，topics 候选 21458	规模已偏大，但 8% 来自 uckkk 单一批量号，真实规模需去水。
增长	weeklyDiff 相对 2026-10-11 基线累计 +935；byWeek 2026-10-12 为 592	周新增从 8 月 4062 峰值回落至 592，生长速度明显放缓。
质量分布	S 108、A 540、B 7402、C 5462、D 2169；avgScore 73.7	S/A 占 4.1%，C/D 占 48.7%，中位质量一般，尾部风险高。
活跃度	active7 2756（17.6%）、active30 6933（44.2%）	近 7 日活跃不足两成，多数插件为“死代码”，生态参与度低。
npm 发布率	published 6901（44%），unpublished 8780，stale 4160	超过一半插件未发布 npm，用户侧安装与更新链路薄弱。
文档覆盖	readme 15641（99.7%），zh 7557（48.1%），none 40	文档基础覆盖好，但中文文档仅约一半，生态国际化程度一般。

cohortTrend：2026-08-10 至 2026-10-12，新插件队列平均分从 74.5 升至 78.7；但队列规模从 3969 降至 592，说明平均分提升部分源于小样本筛选，不能简单解读为生态整体变优。

异常与应对

owner-cluster (mid)

可能原因（假设）：账号 uckkk 使用脚本批量注册/模板复制，生成 1286 个低质插件以占据生态长尾与搜索曝光。
影响面：全生态规模、周增长、作者榜被严重污染，真实质量与活跃度被稀释。
应对建议：DSH 官方对 owner 维度实施插件数/均分双重熔断，下架或隔离 uckkk 及派生插件；报告中单列去水后的核心指标。

owner-cluster (mid)

可能原因（假设）：账号 CAI-MH 属于小规模批量复制或模板填充，9 个插件均分 57 且 1 个无 README。
影响面：范围有限，但提示可能存在更多未被单一阈值捕获的小型批量号。
应对建议：将 CAI-MH 加入观察名单；对缺 README 插件强制补全否则降低可发现性。

dominant-owner (low)

可能原因（假设）：与 uckkk 同一账号长期批量生成未清理，形成单一作者占 8% 的极端主导。
影响面：统计数字虚高，活跃率、增长率失真，误导作者判断生态繁荣度。
应对建议：DSH 官方建立作者信誉分并展示在插件页；用户搜索时过滤批量账号内容。

npm-stale (low)

可能原因（假设）：发布流程未自动化，仓库更新但未同步 npm；部分 npm 包可能被抢注或错配。
影响面：npm 安装用户无法获得最新修复，尤其 dsh-market、dsh-web 等关键插件版本落后，可能遇到已修复 bug。
应对建议：下载页显式展示 repo vs npm 版本差异；核心插件启用 CI 自动发布，作者核查 npm token。

breaking-release (high)

可能原因（假设）：官方在 v0.2.1-rc.1 与 alpha.3 中调整 API/接口，引入 breaking 变更。
影响面：本周 5 个发布中 2 个 breaking，插件作者需要快速适配；未及时适配将导致插件运行时错误。
应对建议：官方发布 breaking 变更公告与迁移脚本；插件作者立即运行兼容性测试并关注 exports["./client"] 等高频缺陷。

下一阶段重点与建议

插件作者：优先检查 dsh-v0.2.1-rc.2 下的 manifest exports 与 client 入口（topDownloads 前 10 中 8 个缺失），并接入 npm 自动发布。
插件作者：避免继续堆叠 whale-widget 二创，转向 remote-mobile、terminal-ssh、memory-knowledge 等低密度场景，每个目前仅 6 个插件。
生态用户：安装前查看 grade 与 npm 版本漂移，不要仅按周下载量选择；警惕 C 级低质高用插件（如 schedule、dsh-cost-meter、dsh-wechat）。
生态用户：用 dsh-whale-report、dsh-git 等 S 级宝藏替代功能模糊的同类插件，并通过 star/评价反哺高质量作者。
DSH 官方：对 uckkk 立即限流/下架，恢复作者榜可信度；发布 breaking 迁移指南，提供 dsh compat check 命令。
DSH 官方：将 quadrant.gems 中的 S 级低星插件加入官方精选列表，并提升 LLM 能力标注覆盖率（当前 11.6%），改善可发现性。

官方动态解读

dsh 核心仓库 253154 stars，最新发布 dsh-v0.2.1-rc.2（2026-10-17）：本周 5 个版本，其中 2 个含 breaking，官方迭代速度提升但适配风险同步上升。
平台生态项目强劲：DeepSeek-V3 104508 stars、DeepSeek-R1 91894 stars、DeepSeek-OCR 24024 stars，为插件提供模型与 OCR 等底层能力，但插件侧更热衷于“免登录/破甲”而非官方 API 合规接入。
官方对市场治理尚无公开动作：本周无插件下架/精选变更记录，仅 weeklyDiff 中 downgrades 为空；建议官方在下周针对 uckkk 批量号和 breaking 发布给出明确治理或迁移信号。

DSH Ecosystem Insights · 2026-W42（2026/10/12～2026/10/18）

Generated by the DSH Insights pipeline + AI · snapshot 2026-10-19 · heuristic evaluation, not a security audit

Core Conclusions

Scale inflation vs real engagement：Out of 15681 authoritative plugins, only 17.6% are 7-day active and 44.2% 30-day active; a single account uckkk owns 1286 plugins with avg score 56, so size must be de-watered before interpretation.
Median quality is barely healthy, tail risk persists：avgScore 73.7, S/A combined only 648 (4.1%), C/D at 48.7%; topDownloads top 10 still includes 3 C-grade low-quality-high-usage plugins.
Official breaking releases plus npm drift create dual adaptation pressure：dsh shipped 5 releases this week, 2 with breaking changes; 10 plugins have repo versions ahead of npm.
Download traffic and quality gems are severely disconnected：billion-context is top download but only B-grade 86; quadrant.gems has 8 S-grade 100-score plugins all with stars <30.
The “status/monitoring/usage” lane is flooded with speculative clones：dsh-whale-widget and several low-quality derivatives occupy 5 download slots; jailbreak tools like dsh-purge have high stars but no npm publish.

Head Plugin Deep Dive

Ebony-Vinyl/dsh-our-free-model：No-login/Key model call for DeepSeek V4.1 Flash / Kimi K3, 24760 stars, A-grade 93, weekly 7921; its actual weekly downloads are less than 1/14 of dsh-market, indicating hype traffic over sustained use.
dsh-market/dsh-market：Official plugin market entrance, A-grade 90, weekly 117724, download rank #3; but npm is at 1.66.42 while repo is 1.43.0, causing version drift—users installing from npm get an outdated market.
ranxianglei/billion-context：Universal context compression proxy, B-grade 86, weekly 450525, #1 by downloads; stars only 1879, inverted traffic vs reputation, likely a hard requirement or recommendation amplification—needs official verification.
MeteorNOX/DeepSeek-Balance-Whale-Widget：Balance monitoring whale, A-grade 90, weekly 55242; it spawned at least two low-quality derivatives (621758/dsh-whale-widget-pet-feed C-81, shixun926-dotcom/dsh-whale-widget-desktop C-71)—original quality but derivative governance needed.
YuJunZhiXue/dsh-purge：Jailbreak tool, 7078 stars, B-grade 78, unpublished npm; weeklyDiff shows +3785 star growth, but compliance risk and no npm publish form a double-edged sword.

Author Landscape

9058 total authors, 2270 multi-plugin; top 10 authors hold only 2.9% star share, yet Ebony-Vinyl alone has 24862 stars—ecology driven by a few star plugins, not an author matrix.
High-quality authors like Fish-under-sea (4 plugins avg 96.5, 13 stars) and EasyTZ (4 plugins avg 91.8, 1 star) have almost no exposure; prolific xby-skill published 175 plugins with only 2 total stars and avg 62.2.
uckkk detected with 1286 plugins, avg 56, 8% of ecosystem, directly pollutes author rankings and growth stats; CAI-MH small cluster also suspected template farming.
Connector nodes Grivn, qikairo7, Jstn-1g combine multiple plugins with high stars (qikairo7 3 plugins 7881 stars), but overall network is sparse, collaboration concentrated in a few individuals.
Judgment: this ecosystem is currently traffic-driven, with nascent community—star plugins attract users, batch accounts inflate scale, and real high-quality contributors lack traffic support.

Quality × Adoption Cross Findings

Underrated gems：SenmuuuuW/dsh-whale-report (S 100, weekly 240) generates daily/weekly reports from session logs—high value; EasyTZ/dsh-git (S 100, weekly 51) provides Git panel, a developer workflow need; official should include and give search slots.
Underrated gems：sunzhentao/dsh--prompt--enhance (S 100, weekly 599) focuses on prompt enhancement in model/proxy lane; Fish-under-sea/dsh-agent-teams-fish (S 100, weekly 1663) is a complementary AgentTeams fork with only 4 stars.
Low quality high usage risks：Bald0Wang/dsh-imggenerate (C 62, weekly 6040) lacks tests/CI/license, users face maintenance or security risks; LuniteGlaze/dsh-plugin-catalog (C 64, weekly 2686) has empty description, opaque functionality.
Low quality high usage risks：ben7am1n/dsh-browser and duyefeng/dsh-browser (C 60-61, weekly 775) are same-name browser automation, intense competition and low quality, users can be misled by download counts.

Ecosystem Theme Map

Theme 1: Context & Memory：session (95), memory (68) tags stand out, billion-context with 450k weekly downloads and dsh-context (A 90) are representatives—context compression/memory is a hard need.
Theme 2: Search & Workspace Enhancement：search (85), ui (73), sidebar (49) form another mainstream, dsh-market and dsh-better-sidebar (S 95) are reps; official scenarios global-search and sidebar-workspace each have only 6 plugins, not yet crowded.
Theme 3: Model Access & Agent Orchestration：model/agent/mcp/skills tags total ~250, MichengAI/dsh-skills-manager (B 88) and dsh-inherit (A 94) manage Agent Skills; but quality variance is large.
Theme 4: Usage & Balance Monitoring：balance (50), usage (50) tags centered, dsh-whale-widget and derivatives are majority, but many clones lower average quality—crowded but unimpressive.
Whitespace opportunities：remote-mobile, terminal-ssh, memory-knowledge each have only 6 plugins; LLM capability tagging covers only 1823/15681 (11.6%), leaving many undiscovered—official should improve tagging and encourage these low-density directions.

Ecosystem Health Assessment

Dimension	Current Value	Interpretation
Scale	authoritative 15681, topic candidates 21458	Too large; 8% from uckkk single batch account, real scale requires de-watering.
Growth	weeklyDiff +935 vs baseline 2026-10-11; byWeek 2026-10-12 is 592	Weekly new plugins fell from Aug peak 4062 to 592, growth significantly slowing.
Quality Distribution	S 108, A 540, B 7402, C 5462, D 2169; avgScore 73.7	S/A 4.1%, C/D 48.7%, median quality mediocre, high tail risk.
Activity	active7 2756 (17.6%), active30 6933 (44.2%)	Less than 20% 7-day active, most plugins are “dead code”, low engagement.
npm Publish Rate	published 6901 (44%), unpublished 8780, stale 4160	Over half not published to npm, weak install/update chain.
Docs Coverage	readme 15641 (99.7%), zh 7557 (48.1%), none 40	Basic docs good, but Chinese docs only half, limited internationalization.

cohortTrend：From 2026-08-10 to 2026-10-12, new plugin cohort avg score rose from 74.5 to 78.7; but cohort size dropped from 3969 to 592, meaning average improvement partly due to smaller sample selection, not necessarily overall ecosystem quality improvement.

Anomalies and Responses

owner-cluster (mid)

Possible cause (hypothesis)：Account uckkk uses scripts for bulk registration/template copying, generating 1286 low-quality plugins to occupy long-tail and search exposure.
Impact：Ecosystem scale, weekly growth, and author rankings heavily polluted; real quality and activity diluted.
Recommendation：DSH official implement dual thresholds on owner plugin count/avg score, remove or isolate uckkk and derivatives; report de-watered core metrics separately.

owner-cluster (mid)

Possible cause (hypothesis)：Account CAI-MH is a small-scale batch copy or template filling, 9 plugins avg 57 and 1 without README.
Impact：Limited scope, but indicates more small batch accounts may exist below thresholds.
Recommendation：Add CAI-MH to watchlist; enforce README completion or reduce discoverability.

dominant-owner (low)

Possible cause (hypothesis)：Same account as uckkk, long-term batch generation not cleaned, creating extreme 8% dominance.
Impact：Inflated statistics, active rate and growth distorted, misleading authors about ecosystem prosperity.
Recommendation：DSH official establish author reputation score and display on plugin pages; filter batch account content in search.

npm-stale (low)

Possible cause (hypothesis)：Publishing not automated, repo updated but npm not synced; some npm packages may be squatted or misconfigured.
Impact：npm installers cannot get latest fixes; critical plugins like dsh-market, dsh-web lag behind, may hit already-fixed bugs.
Recommendation：Download page explicitly show repo vs npm version diff; enable CI auto-publish for core plugins; authors verify npm token.

breaking-release (high)

Possible cause (hypothesis)：Official adjusted APIs/interfaces in v0.2.1-rc.1 and alpha.3, introducing breaking changes.
Impact：5 releases this week, 2 breaking, plugin authors must adapt quickly; failure leads to runtime errors.
Recommendation：Official release breaking change notice and migration scripts; plugin authors run compatibility tests and watch high-frequency defects like missing exports["./client"].

Next Steps and Recommendations

Plugin authors：Prioritize checking manifest exports and client entry under dsh-v0.2.1-rc.2 (topDownloads top 10 has 8 missing), and integrate npm auto-publish.
Plugin authors：Avoid continuing whale-widget clones; pivot to low-density scenarios remote-mobile, terminal-ssh, memory-knowledge (each only 6 plugins).
Ecosystem users：Check grade and npm drift before install; don't choose purely by download count; beware C-grade high-usage plugins like schedule, dsh-cost-meter, dsh-wechat.
Ecosystem users：Use S-grade gems like dsh-whale-report, dsh-git instead of vague alternatives, and star/review to support high-quality authors.
DSH official：Immediately rate-limit/remove uckkk, restore author ranking credibility; publish breaking migration guide and provide dsh compat check command.
DSH official：Add S-grade low-star gems from quadrant.gems to official featured list, and increase LLM capability tagging coverage (currently 11.6%) to improve discoverability.

Official Dynamics

dsh core repo 253154 stars, latest release dsh-v0.2.1-rc.2 (2026-10-17)：5 releases this week, 2 breaking; official iteration speed up but adaptation risk rises.
Platform projects strong：DeepSeek-V3 104508 stars, DeepSeek-R1 91894 stars, DeepSeek-OCR 24024 stars provide model and OCR capabilities, but plugin side favors “no-login/jailbreak” over official API compliance.
No public market governance yet：No plugin removal/feature changes this week, weeklyDiff downgrades empty; official should respond to uckkk batch account and breaking releases next week with clear governance or migration signals.