A DSH plugin with a sidebar workbench and two agent tools that push SKU cut-outs into PSD product-image templates and export a whole batch of JPG/PNG (or ZIP) locally.
DSH integration
Native runtime
Author-claimed
Safety audit
Unaudited
Last verified
2026-11-15
License
NOASSERTION
01What can it help you accomplish?
Turn one PSD main-image template plus a folder of SKU cut-outs into a full batch of product images
One JPG/PNG per SKU at the template's original resolution, optionally packed into a ZIP
E-commerce design and operations teams that already own PSD main-image templates and need many variants
Let an agent render an entire SKU directory unattended on the Host side
JPG/PNG (optional ZIP) written into the session workspace, plus the output directory and the first 50 filenames
Operations and automation teams running overnight batches or product-catalog integrations
Move batch production from designers to operations without a design queue
Operations, agencies or store managers produce the export themselves from the designer's PSD and slot rules
E-commerce operations, agencies and store managers who receive templates from a design team
02How to install into DeepSeek Harness
Prerequisites
- DeepSeek Harness with the plugin manager, dsh `0.2.0-rc.2` or newer (`dsh.engines.dsh >= 0.2.0-rc.2`)
- Node.js 20+ (needed for builds and the Host-side agent tools)
- npm registry access at install time — the agent tools depend on the platform prebuilt binaries of `@napi-rs/canvas`
- No Photoshop installation required
Installation steps
- 01
In the plugin manager run `plugin_manager install_bundle, target = dsh-psd-batch-studio` (the README recommends the npm source; pin a version with `dsh-psd-batch-studio@0.2.6`)
- 02
Or install from GitHub with `plugin_manager install_bundle, target = github:vinkwork/dsh-psd-batch-studio` — pnpm locks the dependency to the commit present at install time, so reinstall after the repository changes
- 03
If the plugin manager does not accept a GitHub spec, run `git clone https://github.com/vinkwork/dsh-psd-batch-studio.git`, `npm install` in the clone, then `plugin_manager install_bundle, target = /absolute/path/dsh-psd-batch-studio`
$ git clone https://github.com/vinkwork/dsh-psd-batch-studio.git
- 04
Refresh the Harness page after installing
- 05
`client.js` / `server.js` are committed with the repository, so no build is needed before installing; run `npm run build` only after changing the sources
$ npm run build
Verify the integration
- The sidebar shows the 「主图批量替换」 (batch main-image replacement) icon
- Two tools appear on the agent side: `psd_inspect` and `psd_batch_render`
- Opening the workbench from the sidebar and loading a `.psd` / `.psb` shows a layer tree on the left with auto-detected product slots outlined by dashed boxes
03DSH integration and capability boundaries
Native DSH plugin bundle installed through the plugin manager (`plugin_manager install_bundle`) — a sidebar workbench page plus two Host-side agent tools, `psd_inspect` and `psd_batch_render`
Template workbench for `.psd` / `.psb`
A PSD/PSB main-image template→Layer tree with per-layer visibility toggles and the live composite on an interactive canvas
Product slot detection and product clusters
Layer tree plus area, position, aspect-ratio and naming rules→Auto-recommended product slots that can also be marked or unmarked by hand
A product's original layer plus its shadow / highlight copies form one cluster: they hide together and the product is drawn onceBatch fitting and canvas fine-tuning
Multiple PNG / JPG / WebP SKU cut-outs→Fit modes (contain / by height / by width / cover with crop), alignment and fill ratio, plus offset / scale / angle per SKU
Current fine-tuning can be copied to every SKU in one clickBatch export
Fitted product slots and the imported SKU queue→JPG or PNG per SKU at the template's original resolution, optionally packed as a ZIP
Agent tools `psd_inspect` / `psd_batch_render`
Template path, SKU directory, output directory, format / fit / anchor and per-SKU adjustments→`psd_inspect` returns template size, layer tree (names / hierarchy / visibility / bounds / layer ids), auto-detected slots and text-layer styles; `psd_batch_render` returns the rendered files plus a `suppressed` list
Both tools resolve paths relative to the current session working directory and reject out-of-bounds paths`psd_batch_render` defaults to JPG at quality 92 and returns the output directory plus the first 50 filenames
04Who is it for? When not to use it?
Good for
- E-commerce design and operations teams that already own PSD main-image templates and need many variants
- Operations and automation teams running overnight batches or product-catalog integrations
- E-commerce operations, agencies and store managers who receive templates from a design team
Not for
- The license is a modified Apache License 2.0 with extra conditions: commercial use is allowed, including internal use and use as a backend or enterprise application platform, but running a multi-tenant service or removing/changing the front-end copyright, project name or Genca/VinkWork attribution needs written commercial authorization from Shanghai Genca; versions up to v0.1.2 were released under PolyForm Noncommercial 1.0.0.
- Version 0.2.5 is unusable: the README states its agent tools cannot register on any dsh version (issue #7), so if `psd_inspect` / `psd_batch_render` are missing after installing, check the version first — the npm target to pin is 0.2.6. Also never mount two copies of the plugin in one profile, because both use the same `psd-batch-studio` entry.
- Text layers are composited from the raster pixels stored in the PSD, so copy and fonts cannot be edited through the tools; bevel/emboss and inner shadow are approximated or skipped and stroke / drop shadow / glow are approximated. One template is processed per run (the SKU queue persists when switching templates), a multi-hundred-MB PSB needs hundreds of MB of RAM, and the agent tools need a loadable `@napi-rs/canvas` native build — the in-page workbench still works where that native package fails.
05Compatibility, maintenance and safety notes
- The license is a modified Apache License 2.0 with extra conditions: commercial use is allowed, including internal use and use as a backend or enterprise application platform, but running a multi-tenant service or removing/changing the front-end copyright, project name or Genca/VinkWork attribution needs written commercial authorization from Shanghai Genca; versions up to v0.1.2 were released under PolyForm Noncommercial 1.0.0.
- Version 0.2.5 is unusable: the README states its agent tools cannot register on any dsh version (issue #7), so if `psd_inspect` / `psd_batch_render` are missing after installing, check the version first — the npm target to pin is 0.2.6. Also never mount two copies of the plugin in one profile, because both use the same `psd-batch-studio` entry.
- Text layers are composited from the raster pixels stored in the PSD, so copy and fonts cannot be edited through the tools; bevel/emboss and inner shadow are approximated or skipped and stroke / drop shadow / glow are approximated. One template is processed per run (the SKU queue persists when switching templates), a multi-hundred-MB PSB needs hundreds of MB of RAM, and the agent tools need a loadable `@napi-rs/canvas` native build — the in-page workbench still works where that native package fails.
NOASSERTION · latest release v0.2.5 (2026-11-06), which the README marks as unusable — npm target to pin is 0.2.6 · last push 2026-11-13
06Frequently asked questions
How do I install DSH PSD Batch Studio in DeepSeek Harness?
Run `plugin_manager install_bundle, target = dsh-psd-batch-studio` in the plugin manager — that is the npm source the README recommends, and a version can be pinned with `dsh-psd-batch-studio@0.2.6`. Alternatively use `target = github:vinkwork/dsh-psd-batch-studio`, or clone the repo, run `npm install` and install from the absolute local path. Refresh the Harness page afterwards: the 「主图批量替换」 icon appears in the sidebar and two tools appear on the agent side.
The plugin installed but `psd_inspect` / `psd_batch_render` are missing from my agent session. What now?
Check the version first: 0.2.5 fails dsh's tool-schema validation on every dsh version (issue #7) and must not be used. Also make sure the same profile does not carry two copies of the plugin — for example an old local install plus the GitHub build — because both share the single `psd-batch-studio` entry.
Install fails with `ERR_PNPM_UNEXPECTED_STORE` — is that the plugin?
No, it is a local environment issue: the profile's `node_modules` was previously installed by a different pnpm major (pnpm 10 in a terminal versus the pnpm 11 bundled with the Harness desktop app) and the stores cannot be mixed. Move `~/.dsh/profiles/<profile>/node_modules` aside as a backup and reinstall inside Harness.
Do my PSD templates and SKU images leave the machine?
No. PSD parsing and compositing run locally, either in the page or in the Host process, and the README states there are no network requests, no telemetry and no asset uploads. The only network requirement is npm registry access at install time, because the agent tools depend on platform prebuilt packages of `@napi-rs/canvas`.
What does a batch run produce, and do I still need Photoshop?
Photoshop is not required. A run outputs one JPG or PNG per SKU at the template's original resolution, optionally packed into a ZIP; `psd_batch_render` defaults to JPG quality 92 and returns the output directory plus the first 50 filenames. Text layers are composited from the PSD's raster pixels, so copy and fonts cannot be changed from the tools, and one template is processed per run — the SKU queue is kept when you switch templates.
07Related DSH workflows
dsh-image-gen
by shanliuling
Adds in-chat multimodal image generation via a `generate_image` tool with multi-provider support (Gemini/OpenAI/Seedream), interactive preview, copy/download, and session-attached results.
dsh-vision-router
by ysr666
Eyes for text-only DeepSeek Harness agents: built-in free vision chain (no key) + pixel-level vision tools (Q&A, grounding, crop, pixel diff, colors, OCR, SVG trace, cutout, screenshots). One-command install, no Python, image turns work like ordinary tool-calling turns.
dsh-vision-toolkit
by anionex
为纯文本 DSH Agent 提供 10 个结构化视觉工具:意图感知图片问答、长截图 OCR、原始像素 grounding、UI 还原、像素 diff 等
dsh-see-now
by saos-3624
Captures screen, OBS, or webcam frames, filters by visual change, packs them into image sheets, and delivers them to DSH sessions.
08Data and sources
一个 [DeepSeek Harness](https://github.com/deepseek-ai) 插件,用来批量生产电商商品主图:
- DeepSeek Harness(带插件管理器),dsh `0.2.0-rc.2` 及以上(`dsh.engines.dsh >= 0.2.0-rc.2`)。
安装后刷新 Harness 页面:侧边栏出现「主图批量替换」图标,agent 侧出现两个工具。
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-11-15. Found an error? Submit a correction.
Best DeepSeek Harness Plugins
Twelve plugins worth installing first — picked from the whole catalog, across every category.
