Stringers
Roland0511/dsh-marktext · Roland0511

13 minutes, 24 forks: a DSH Markdown editor plugin. How many of its 9-day 458 stars are real?

What it is A DeepSeek Harness sidebar plugin: open a .md file and edit it WYSIWYG, quote selected passages into the chat, or export to PDF in one click.

Why it took off DSH's built-in sidebar can only preview Markdown. This plugin borrows MarkText's mature core for a full editing experience. The author submitted it to various awesome lists the day the repo was created and got 86 stars in the first 5 days. Later, stars surged while npm downloads declined, and the forks came from a burst of newly registered alt accounts within 13 minutes. Likely botted.

CHow you could follow it Don't build another DSH Markdown editor. Borrow the approach: plug a proven open-source engine into the extension slot of a new host, fill the gap of "can view but not edit, and edits can't go back to the agent," get listed on awesome lists the day the repo goes up, and pair it with a demo video.

C grade: the approach is worth learning, but this spot already has dozens of similar plugins, and the closest same-function rival has only 13 stars. Removing the botted stars, the real result is roughly 86 stars in 5 days. Copying it will most likely get you double-digit stars. Few people know about it yet
Data (updated daily) · Repo created 2026-11-02 · 846 stars · 7 days +295
GitHub · stars
05001k11-0411-0811-1211-1611-20846
Data 2026-11-04 → 2026-11-20 (17 days, 17 observations). Points before 10-28 come from deeply / archive backfill, not every day has one

dsh-marktext moves the core of the veteran editor MarkText into the right sidebar of DeepSeek Harness (DSH). Open a .md file there and edit directly, then quote selected passages back into the chat. The gap it fills is real, and it's done with care. But its star count, 458 after 9 days, needs to be read in two parts: the first 86 stars in the first 5 days came from the author submitting it to awesome lists, while the large chunk after that doesn't match real install numbers, and the fork pattern already shows signs of batch-created alt accounts. Below: what it does, how the first 86 stars came in, why the later numbers should be discounted, and what a DSH plugin hoping to earn stars can learn from this case.

Nov 8 at 2 PM, 24 forks in 13 minutes

2026 Nov 8 Beijing time 13:55 to 14:08, the repo Roland0511/dsh-marktext was forked 24 times in 13 minutes, with 6 of them in 14:02 alone. Another batch came more than an hour later. The repo's fork count went from 17 the previous day to 48 (checked per entry via the GitHub forks API, by creation time).

Looking at these accounts, they all look alike. liu-qiangaem has the display name "Zoe Myers", registered Oct 27, 0 followers, and 5 repos total: one called comfy-hill-tool, one forked echarts, plus dsh-marktext. ma-chaorz has the display name "Joran Costa Sofia", registered Oct 26, with repos lucky-atlas-881 and quiet-vessel-882. Username and display name don't match, repo names are random words plus numbers, and they fork a few famous repos like ComfyUI and opencv to look legit.

Of the 54 fork accounts in total, 47 were registered in October, 2 in Nov 2, 4 have already been banned or deleted by GitHub, and only 1 are older accounts. Comparing against 4 new DSH plugins with similar star counts from the same period, these accounts barely forked anything else. They only went for this one repo.

Who bought these forks can't be found from outside. It could be the author, or a botting service that picked a repo that was already trending as a showcase. GitHub doesn't show star-level details, so whether the stars were botted can't be confirmed. It can only be inferred from the surrounding evidence. But before getting to that part, one thing should be clear: the plugin itself is seriously built.

What it is: edit the documents an agent writes, right in DSH's sidebar

DSH is an agent tool open-sourced by DeepSeek in August, and its plugin ecosystem is hot. When agents work, they produce lots of .md files, like plans, reports and notes. The official right sidebar can preview Markdown but not edit it, so to change anything you have to switch to Typora or VS Code.

dsh-marktext fills exactly that gap. The first line of the README reads "Write Markdown the way it will look." It does several things:

  • WYSIWYG: headings, tables, formulas and charts are edited in place. Click into a table cell to edit that cell, no need to edit source code with a wall of pipes.
  • Quote into chat: select a passage and hit Quote, and a quote block with the file path is appended to the session's input box. After editing a document the agent wrote, you can feed a specific note straight back. Among similar plugins, this step is closest to how agents are actually used.
  • Export PDF: press ⌘P to generate an A4 PDF saved next to the .md file, with no print dialog.
  • Safe saving: auto-saves 2.5 seconds after you stop typing. If someone else changed the file on disk, a conflict prompt appears instead of silently overwriting it.
  • KaTeX, Mermaid and PlantUML are bundled into the plugin for offline rendering, and light and dark themes follow the host.

The editor itself wasn't written by the author. He used muya, the open-source core split out of MarkText (MIT license), and only wrote the layer that plugs it into DSH. So it shipped with full tables, formulas and charts from day one. The engineering is tidy too: there are tests, releases are fully automated, and the README is bilingual (Chinese and English) and reads like a product page, with motivation, FAQ and known limitations. Nov 2 the repo was created and he shipped 9 versions that same day; by 11-07 there were 13 in total.

The author Roland0511's bio says "game developer." He joined GitHub in 2011, has 13 followers, previously contributed a PR to langchainjs, and has no traffic channel of his own.

First 5 days' 86 stars: lining up on every list the day it launched

On the day the repo was created (11-02), the author submitted it to awesome-dsh-plugin, and it was later merged. 11-07 he then submitted PRs to 3 more awesome lists in one go, and 1 were merged. After that, a series of auto-generated aggregator sites picked it up: dsh-plugin.org rewrote the README into a detail page, along with Trendshift, Oh-My-DSH, dshcrate, and several awesome lists sorted by stars.

DSH's plugin install wizard lets you install by entering the package name, and dsh-marktext is a very straightforward name to search for. npm downloads on day one were 1,634 (9 versions that day; mirror syncing inflates the count), and in the following days they were 400, 318, 114 and 83 (npm download stats).

Stars over these days were 44, 66 and 86 (from 11-04 to 11-06, per-day star counts from GitHub). A plugin with no personal traffic, 0 discussions on X, and no media or official endorsement, relying on polish and active list submissions, reached 86 stars in 5 days. This stretch is believable.

After 11-07: stars accelerate while installs decline

Datestarforknpm downloads
11-0686683
11-0720313228
11-0826317102
11-092974895
11-1035650—
11-1145850—

(Stars and forks are daily GitHub counts; npm figures come from npm download stats, and the 11-10 figure was not yet out that day.)

On 11-07 the author pushed to several lists at once, so a jump to 117 stars is somewhat explainable. After that, stars added thirty to a hundred per day, while npm downloads fell back to around 100 per day. The fork spike came in exactly the 13 minutes of 11-08.

The most direct comparison is another plugin in the same slot: dsh-markdown-typora (by ltmroberthk915, the markpilot project). It also edits .md files in the right sidebar, also has overwrite-protected saving, and adds draft recovery on top. On 11-07 and 11-08 its npm downloads were 884 and 500, more than dsh-marktext's 228 and 102 on the same two days, totaling 2,362 over 30 days versus dsh-marktext's 2,974. Install numbers are roughly similar, yet dsh-markdown-typora's repo has only 13 stars, 35 times fewer.

With similar numbers of users but such a gap in stars, only two explanations fit: dsh-marktext got aggressive list placement, and the later stretch was most likely botted (an inference). Aggregator sites mostly sort by stars, so more stars means higher placement, and higher placement brings more stars, and the two amplify each other. It once ranked #25 on Trendshift's JavaScript chart.

This spot is already crowded

ProjectScaleCompared with alternatives
DSH's official right-sidebar Markdown previewBuilt into the hostView only, no editing; if the official side adds editing someday, this category loses its place
omdsh-dev/DSH-better-sidebar5,057 starsFull sidebar bundle with built-in file editing; dsh-marktext only does Markdown but goes deeper
dsh-markdown-typora(markpilot)13 stars, 2,362 npm downloads in 30 daysHead-on competitor with comparable installs; both claim the same .md open slot, so users must pick one
dsh-sidebar-editor, two dsh-md-editor0–2 starsRich text plus source dual view, Monaco source editing, etc.
dsh-markdown and dozens more0–45 starsMostly preview or light editing
Typora, MarkText, ObsidianDesktop appsRequires switching windows, and can't send edits back to the agent

(GitHub and npm, as of 11-10.)

Searching "dsh markdown" on GitHub returns dozens of repos, and the dsh-plugin topic has about 30k repos. dsh-marktext is neither the first nor the only editor, and 11-08 and 11-09 have new similar repos being created.

If it were you

C: you can learn from it, but don't pick this topic. Removing the botted part, a plugin this polished that also actively placed on lists has a real result of roughly 86 stars in 5 days, while its closest rival in the same slot has only 13 stars. Building another DSH Markdown editor will most likely get you double-digit stars. What's worth taking away is the method.

Who it suits: developers who want to earn their first starred open-source project in a hot new agent ecosystem and are familiar with a mature open-source engine (editors, whiteboards, spreadsheets, charts).

Who it doesn't suit: anyone wanting to build yet another DSH sidebar Markdown editor; anyone hoping to reach hundreds or thousands of stars just by list placement; anyone starting with botting (alt-account forks in batches every 13 minutes, names that don't match usernames are easy to spot, and npm downloads give it away too).

How to do it:

  1. Borrow a mature engine, don't write from scratch: find a proven open-source engine (muya, CodeMirror, Monaco, Excalidraw, etc.), plug it into the host's extension slot, and you get polish on day one. We didn't check which engines have already been integrated into DSH one by one, so search GitHub before you start.
  2. Look for "can view but not edit" or "can view but not send back": file types the official side already previews, plus editing, plus an action that returns to the agent (like Quote into chat). There are already too many plugins that only render.
  3. Write the README as a product page: explain why you built it, a one-line install, an FAQ and known limitations, automate releases, and push out versions frequently in the first few days.

How to get users: on the day the repo goes up, submit it to every awesome list and plugin aggregator. This is the real starting volume for projects like this. Use a straightforward package name people will search for. List placement alone gets you a few dozen stars. In the same ecosystem, dsh-see-now took off through a single Bilibili demo video (350k views bringing about 530 stars), which shows that for more, you need a demo video that clearly explains "what it saves you."

When to move: DSH plugins probably still have a few months of novelty left (an inference), and new repos appear every day. Ideally ship before Q1 of 2027.

What would mean I misread it: ① Once the stargazer breakdown is accessible and shows dsh-marktext's stars mostly come from old accounts, the second half is real too, and chart-stuffing mattered more than I estimated here. ② If someone reached 300+ stars on a single DSH plugin purely through chart listings, with no videos and no botting, that would mean the aggregator-list route is underrated. ③ If the DSH official team adds Markdown editing to the right sidebar, this slot gets taken.

GitHub repo ↗

Research notes: competitor table, acquisition evidence, full research report
Competitors
NameWhereLaunchedScaleHow it differs from this one
dsh official right-sidebar Markdown previewBig-company featuresFrom ¥2026-08Built into the host; every dsh user has itOfficial is read-only preview. This one edits, quotes into the chat, and exports PDF. Once official adds editing, this niche is gone
omdsh-dev/DSH-better-sidebarOpen source2026-08-075,057 stars / 483 forksA sidebar workspace suite with built-in file rendering and editing. This one does only Markdown, but goes deeper
dsh-markdown-typora(ltmroberthk915/markpilot)Open sourcenpm 2026-11-0313 stars; 30 npm downloads over 2,362 daysHead-on overlap: same .md editing in the sidebar, same version-guarded save, built on CodeMirror 6, plus draft recovery. Install base is comparable; star count is 35× higher
DoggyHU/dsh-sidebar-editorOpen source2026-10-211 starsMilkdown rich text plus CodeMirror dual view
yadnuses/dsh-md-editor, Zhang1025681484/dsh-md-editorOpen source2026-10-31 / 10-092 / 0 starsOne pitches edit-in-place with zero drift; the other is source editing with Monaco
yueyezhufeng/dsh-markdown, dsh-markdown-ide, dsh-markdown-reader, dsh-md-preview and dozens moreOpen source2026-08~100–45 starsMostly preview or light editing. None has muya-level WYSIWYG, and none quotes into the chat
Typora / MarkText / Obsidian themselvesOtherYearsMainstream desktop appsYou have to switch out of dsh to use another app. This one stays in the window and can send results back to the agent
Where users come from
  • Verified awesome lists and plugin aggregator sites: the author 11-02 submitted PR #11428 to awesome-dsh-plugin (merged), then 11-07 submitted PRs to imsai-sh (merged), Zhiyuan-Fan, and 0xsline. It was later listed by dsh-plugin.org, Trendshift, Oh-My-DSH, dshworks, cccakeee, vvlife, and dshcrate.
  • Inferred dsh host plugin install wizard / npm search by package name: the package name is literally dsh-marktext, and the README leads with a one-line install. npm holds steady at about 100 downloads per day
  • Confirmed fork boosting: of the 54 fork accounts, 47 were registered on 2026-10, 2 on 11-02, and 4 have since been 404. About 25 forked in a row between 11-08 05:55–06:08 UTC. Usernames and display names don't match (liu-qiangaem's display name is Zoe Myers). 0 followers. Random repo names plus forks of popular repos to age the accounts. Almost no overlap with the forkers of 4 other popular dsh plugins from the same period. This is targeted boosting
  • Inferred star boosting: after 11-07 stars accelerated (+117, +102/day) while npm downloads fell to about 100/day over the same stretch. The stargazer breakdown isn't accessible from this environment

Why dsh-marktext (Roland0511/dsh-marktext) got traction

Channel: GitHub | Research date: 2026-11-10 | Data card: repo created 2026-11-02,9 days ago, 458 stars / 50 forks, +7 stars in the last 414 days. GitHub API live check that day: 459 stars / 50 forks

Bottom line

It's a thoroughly built DeepSeek Harness (dsh) plugin: it drops MarkText's editor core, muya, into dsh's right sidebar so you can edit .md WYSIWYG, quote selected paragraphs into the chat box, and export to PDF. It fills a real gap: dsh officially can only preview Markdown, not edit it. But take the "9 days, 458 stars" figure with a grain of salt. For the first 5 days, the plugin climbed organically to 86 stars via aggregator lists, but the star acceleration after that doesn't match real installs: npm downloads fell from 1,634 on day one to about 100 a day. Of the 50 forks, 53/54 came from 2026-10-26~11-02 accounts registered just before, clustered within 13 minutes around 11-08 06:00 UTC (another 4 accounts have since been banned or deleted by GitHub). Bottom line: the product has real niche demand and polish, but most of the "breakout" star count is probably fake. Fork inflation has evidence behind it; star inflation can't be verified without stargazer data, so it's unconfirmed. Who did it is also unconfirmed.

1. What it is

  • A third-party dsh plugin published on npm as dsh-marktext. Install with one line: dsh plugin --profile web add dsh-marktext. In the desktop app, enter the package name in the Add plugin wizard (per the README).
  • The editor core is @muyajs/core (muya, extracted from MarkText, MIT). Only the dsh integration layer is custom. The editor runs in an iframe to isolate styles.
  • Selling points (README): ① WYSIWYG editing; click into a table to edit a cell. ② Select text and click Quote, and a file-path citation block is appended to the draft in the input box of "the session this file belongs to." ③ ⌘P exports A4 PDF saved next to the .md. ④ KaTeX / Mermaid / Vega / PlantUML render offline. ⑤ Saves are version-guarded: a conflict shows a banner instead of silently overwriting. ⑥ Follows the host's light and dark themes.
  • Execution is strong: tests, a headless Chrome verification script, automated releases via npm Trusted Publishing (OIDC), bilingual Chinese and English READMEs, plus a Star History chart. 11-02–11-07 it shipped 13 versions (v0.1.1→v0.1.13). 11-02 alone saw 9 releases (GitHub releases; 15 total versions on the npm registry).
  • Author lizhou.zhu (GitHub Roland0511, bio "game developer", account created 2011, 13 followers, 48 public repos, mostly forks). Earlier contributed PRs to langchainjs (Milvus vectorStore) and hybridclr, and in October built a macOS utility, codex-pulse (1 stars). No personal audience. The X account @lizhou_zhu returns nothing related in the x-radar corpus.

2. Growth data and timeline

TimelineEvent / numberSource
2026-11-02 08:45Repo created; 11:06 npm first release; 9 versions shipped that dayGitHub API;registry.npmjs.org
2026-11-02Author's PR #11428 to awesome-dsh-plugin listing it, mergedgh search prs
2026-11-02npm downloads 1,634 (9 versions that day; mirror syncs inflate this)api.npmjs.org
2026-11-03 07:523 accounts forked in a row within 14 seconds; all 3 accounts are now 404 (banned or deleted)GitHub forks API
11-04 / 11-05 / 11-0644 / 66 / 86 stars; npm downloads 318 / 114 / 83Daily data card; api.npmjs.org
2026-11-07Released v0.1.13. The same day the author submitted PRs to 3 more awesome lists (imsai-sh merged; Zhiyuan-Fan and 0xsline not merged). Stars went 86→203 (+117, the largest single-day gain in the observed window). npm 228gh search prs; data card; npm
2026-11-08 05:55–06:08 UTCAbout 25 newly registered accounts forked in a row within 13 minutes. Another wave came 08:21–09:17. Forks 17→48GitHub forks API; data card
11-08~11-11Stars 263 → 297 → 356 → 458 (11-11 +102). Over the same stretch npm averaged only 102 / 95 downloads per dayData card; api.npmjs.org
Cumulativenpm over the last 30 days (10-11–11-09): 2,974 downloads totalapi.npmjs.org

Traits of the fork accounts (spot-checked 9; all 54 fork accounts counted):

  • Of 54 fork accounts, 47 were registered in 2026-10,2, 2026-11-02,4 were already 404, and only 1 (oralyalcin, registered 2019 years ago, 391 repos, 136 followers, looks like a bulk fork account) are old accounts.
  • Usernames and display names don't match. For example, liu-qiangaem displays as "Zoe Myers," and ma-chaorz as "Joran Costa Sofia." Each has 0 followers, 0 following, and 4–5 repos.
  • Self-created repo names are random word pairs plus numbers, like trim-atlas-889 and quiet-vessel-882. The accounts were also aged by forking well-known repos such as ComfyUI, opencv, LlamaFactory, nocodb, and echarts.
  • Targeted at this one repo: compared against 4 new dsh plugins with similar star counts from the same period (dsh-media-desk 96 forks, dsh-guizang-design 43, dsh-see-now 58, dsh-game-studio 42). These accounts overlap with dsh-media-desk on only 1 repos, and the other 3 are 0. So this isn't a farm forking popular repos in passing. It's aimed squarely at dsh-marktext.
  • Star breakdown: in this environment the GitHub stargazers REST endpoint returns 404, and GraphQL totalCount is 0. The stargazer list and account creation dates aren't accessible, so whether stars were boosted is unconfirmed. Indirect evidence: stars accelerated after 11-07 while npm installs declined. Also, the suspect column in our GitHub library is empty for this repo: its fork rate is 11%, which doesn't trigger the "zero forks" rule.

Real-usage check: dsh-markdown-typora does the same job (by ltmroberthk915, from the MarkPilot project, built on the CodeMirror 6 editor, published on npm 11-03). Its GitHub repo has only 13 stars. But its npm downloads over the last 30 days were 2,362,11-07 and 11-08 on the two most recent days, 884 and 500, which is higher than dsh-marktext's 228 and 102 for the same two days (api.npmjs.org). So real installs are about even, while the star count differs by 35×.

3. Where users came from (growth tactics)

  1. Plugin aggregators and awesome lists (confirmed, main channel): the author submitted a listing PR to awesome-dsh-plugin the day the repo was created, and 11-07 submitted 3 more. After that, a run of auto-generated aggregator pages picked it up: the dsh-plugin.org detail page (which rewrote the README), Trendshift (repositories/334854,1 like), Oh-My-DSH (375 stars at the time of recording), dshworks/awesome-dsh-plugins (410 stars), cccakeee (297 stars), vvlife (98 stars), fjzzwxp, and dshcrate.com. Most of these pages sort by star count, so the more stars a repo has, the higher it ranks, and the stars and the listings feed each other.
  2. In-host search and install wizard (inferred): from dsh desktop 0.2 onward, the built-in plugin install wizard lets users search directly. The package name dsh-marktext is easy to find. The steady daily npm downloads (about 100) mainly come from there and from the aggregator lists.
  3. Fork boosting (confirmed) / star boosting (unconfirmed): see the section above. The fork push came early (11-08), right after 11-07's star jumps and the author's concentrated awesome-list submissions. It looks aimed at placing the repo near the top of Trendshift's JS chart (card record #25) and the aggregator rankings. Who paid for it can't be determined: the author or a boosting vendor picking its own targets are both possible.
  4. Channels with no coverage: x-radar found dsh-marktext and marktext only 0 times each in 30 days. Jina search turns up nothing on Zhihu, LINUX DO, V2EX, Bilibili, WeChat Official Accounts or media articles, and no endorsement from DSH officially or from 崔添翼. By comparison, the sibling dsh-context got its traction from an official shout-out and reposts by 爱可可 and 智东西.

4. Competitors

CompetitorsWhereLaunchedScaleHow it differs
dsh official right-sidebar Markdown preview (dsh-client-ui-sidebar-markdown, discussion #6056)Big-company featuressince 2026-08, with 0.2.x improvements ongoingBuilt into the host; every dsh user has itRead-only preview (with Mermaid). dsh-marktext can edit, quote into the conversation, and export PDF. If the official side adds editing someday, this plugin loses its place.
omdsh-dev/DSH-better-sidebarOpen source2026-08-075,057 stars / 483 forksA Codex-style sidebar workbench with built-in "file render and edit", an all-in-one suite. dsh-marktext does one thing, Markdown, but goes deeper.
dsh-markdown-typora(ltmroberthk915/markpilot)Open source / npmnpm 2026-11-03, repo 09-0913 stars; 30 npm downloads over 2,362 daysDirect head-to-head: it also edits .md files in the right sidebar, also saves with version guards, plus draft recovery, and supports VS Code / Trae. Both claim the *.md extension slot, and the README says to pick one. Usage is about the same.
DoggyHU/dsh-sidebar-editorOpen source2026-10-211 starsMilkdown (ProseMirror) rich text plus a CodeMirror dual view
yadnuses/dsh-md-editor, Zhang1025681484/dsh-md-editorOpen source10-31 / 10-092 / 0 starsOne pitches "click to edit where you read, zero source drift"; the other edits source in Monaco.
yueyezhufeng/dsh-markdown, minhdevtry/dsh-markdown-ide, s1lencewill/dsh-markdown-reader, poiuyjie/dsh-md-preview, and othersOpen source2026-08~1045 / 5 / 1 / a handful of starsLightweight editors, Notion-style, full-screen readers, in-session md preview. Mostly preview or basic editing.
Typora / MarkText / Obsidian themselvesOthers (desktop apps)YearsThe MarkText main repo has tens of thousands of stars.You have to switch to a separate app. dsh-marktext's value is not switching windows, and it can "quote into the conversation".

Verdict: this slot is already crowded. Searching GitHub for "dsh markdown" turns up dozens of repos, and the dsh-plugin topic holds about 30k repos (data card). It isn't the first, and it isn't the only one that edits.

5. Why it took off

  1. Real need, but small: agents write lots of .md files (plans, reports). People want to read and edit them alongside, then send a passage back to the agent. The official side only previews, so editing means switching to Typora or VS Code. "Select → Quote into conversation" closes the loop of "human edits agent-written doc, feeds feedback back", the point in its category that fits the agent workflow best.
  2. Borrowed a mature core, finished on day one: it doesn't write its own editor, it uses muya directly, so tables, formulas, Mermaid, and PlantUML work from the start. It also handles iframe isolation, version guards, PDF saving, and offline assets. The README reads like a proper product page (motivation, FAQ, known limits, license table). In an ecosystem of 30k repos, mostly AI-generated in one pass, that level of polish is enough to get it into the aggregator lists.
  3. Actively seeded aggregator lists: on the day the repo was created and through 11-07, PRs went to 4 awesome lists in total, and aggregator sites reposted them automatically. This was the main source of about 86 stars in the first 5 days.
  4. The acceleration after 11-07 is most likely star-farming: there's direct evidence of fork farming. Stars accelerated while npm downloads fell, and install counts are comparable to 13-star rivals in the same category. Both suggest a sizable share of the 400 stars don't come from real users.

6. Differentiation

The real differentiators are two: ① it uses true WYSIWYG from MarkText/muya (most peers are preview-only, source editing, or split panes); ② "quote into the current session draft" plus "PDF saved next to the doc", two actions designed for agent workflows.

These differentiators explain why it's better than a pile of 0-star peers: people are willing to install it, with around 100 npm downloads a day. They don't explain "9 stars in 458 days, and 7 +414 in the last few days". The closest rival, dsh-markdown-typora, is functionally similar with comparable installs, yet has only 13 stars. The star gap comes mainly from list seeding and (likely) star-farming, not from product differences.

7. Visibility

It hasn't spread much in niche circles either. Only aggregator lists and auto-generated directory pages picked it up (dsh-plugin.org, Trendshift, Oh-My-DSH, several awesome lists, dshcrate). No big accounts, media, or DSH official endorsement, and 0 posts on X. Copycats: the 50 forks are mostly throwaway accounts, not followers. Around the same time, a batch of repos doing similar things appeared (dsh-md-editor, dsh-sidebar-editor, dsh-markdown-outline created 11-08, dsh-typora-links created 11-09), but calling them copies would be a stretch.

8. What to borrow

  • "A mature open-source core + a new host plugin": when a new ecosystem appears, don't build from scratch. Wire a proven engine (muya, CodeMirror, Monaco, Excalidraw, etc.) into the host's extension points and you get polish on day one.
  • Find the host's gaps, where it can "see but not edit" or "see but not send back", and add an action that returns to the agent, such as Quote into conversation. That differentiates far more than rendering alone.
  • On the day the repo is created, send PRs to every awesome list, use plain search-term package names, write the README like a product page, automate releases with OIDC, and iterate often.
  • Cautionary note: in an ecosystem of aggregator lists sorted by stars, star-farming can push a repo up the charts short term. But npm downloads give it away, and throwaway fork accounts are easy to spot (batch-registered at similar times, usernames that don't match display names, forks clustered within minutes).

9. Uncertain

  • Star-level detail is unavailable (the stargazers endpoint returns 404 in this environment, GraphQL returns 0), so how many of the 458 stars were farmed is unconfirmed.
  • Who bought the fork accounts (the author, or a farming service picking targets) is unconfirmed.
  • There is no public install data for the dsh host. npm downloads include mirror syncs and CI, so treat them as a rough reference only.
  • Whether the official side adds editing to the right-sidebar Markdown view determines how long this plugin lasts.

Sources

This article was written by AI from public data and the data we collect daily. Numbers are sourced, and inferences are labeled.