At a glance
What it does
Sync Cubox bookmarks into DeepSeek Harness, search them locally, and export Markdown or AI daily briefs to Obsidian.
Web Profile
Not declared in supplied evidence
Evidence-verified
Checked Sep 11, 2026, 2:10 PM UTC
Code-evidenced contributions
What it adds to DSH
Provides configuration, sync, status, and collection-query tools for Cubox bookmarks.
Mechanism evidence ↗Adds a web settings panel for Cubox credentials, sync interval, export settings, and AI-brief settings.
Mechanism evidence ↗Before you choose it
dsh-cubox is a DeepSeek Harness web plugin for syncing recent Cubox bookmarks on demand or on a schedule. It caches cards and annotations locally, lets agents query the collection, and can export per-card Markdown plus an LLM-generated daily brief to a chosen local folder.
Best for
DeepSeek Harness users who collect reading links in Cubox and want searchable local notes or an Obsidian-oriented export.
Common tasks
- Sync today's or recent Cubox saves with cubox_sync.
- Search saved cards by keyword, date window, annotation, star, or read state.
- Export each saved card as Markdown with links and annotations.
- Generate a prompt-template-based daily brief when an LLM is configured.
Permissions and data
Requires a Cubox API Extension credential; optional AI briefs require an OpenAI-compatible LLM credential.
Permissions- Reads and may modify Cubox data through the supplied API Extension credential.
- Writes a local cache at ~/.dsh/dsh-cubox-cache.json.
- Writes credentials to ~/.dsh/dsh-cubox.json; the README states mode 0600.
- Writes Markdown exports and AI briefs to the user-selected output folder.
- Caches bookmark cards and annotations locally.
- Sends the configured collection content to the configured LLM service when generating an AI brief.
- Cubox internal /c/api/cli endpoints.
- Optional OpenAI-compatible LLM endpoint; DeepSeek is the documented default.
- Cubox API Extension link.
- Optional LLM API key for AI daily briefs.
Limitations
- The README says collection reads use Cubox internal endpoints rather than the public open API.
- No Harness version range is declared in the supplied evidence.
- Installation and runtime behavior were not executed or independently audited.
- Package contents from the npm tarball were not audited.
What DSHub checked
- Pinned Git source and bundle structure were verified.
- The manifest declares a web client bundle and Node ^22.19.0 or >=24.0.0.
- The repository documents scheduled sync, local caching, query tools, Markdown export, and an optional AI brief.
What DSHub did not check
- Successful installation in a DeepSeek Harness environment.
- Live Cubox API behavior, data modification scope, and scheduled-sync reliability.
- LLM output quality, cost, and behavior with any configured provider.
Pinned install
Install dsh-cubox
This plugin bundle does not have a DSH Plugin install action. Use its source documentation for the delivery method.
Maintainer source
Project README
dsh-cubox
English | 中文
Cubox sync for DeepSeek Harness: scheduled sync of your Cubox collection, with an AI daily brief generated from your own prompt template and written straight into Obsidian — via the same /c/api/cli endpoints the official cubox-cli uses. Agent tools plus a web settings panel.
Features
- Scheduled sync — a timer pulls your latest bookmarks into a local cache (
~/.dsh/dsh-cubox-cache.json) every N minutes (default 60, configurable; 0 disables the timer). Manual sync anytime withcubox_sync. - AI daily brief — write your own prompt template (e.g. "今日收藏简报",
{collection}is replaced with the formatted collection for the sync window: title / source / summary / annotations). When an LLM key is configured, every sync generates a brief into the output dir —今日收藏简报-YYYY-MM-DD.mdfor a 1-day sync,最近N日收藏简报-YYYY-MM-DD.mdfor wider windows (each window gets its own file). - Markdown export — set
outputDirand optionally keep one markdown file per card (frontmatter + title + description + Cubox/original links + annotations, same layout as the official Cubox Obsidian plugin). ToggleexportCardsoff to write only the AI brief. - Query —
cubox_cardsfilters by keyword, time window, annotated/starred/read status. - Config & status —
cubox_config/cubox_status; credentials persist to~/.dsh/dsh-cubox.json(mode 0600), secrets never echoed. - Settings panel — Settings → Cubox: paste the API-extension link, set the sync interval, pick the local export folder (OS folder chooser), toggle per-card export, edit the AI brief prompt and LLM settings (OpenAI-compatible, defaults to DeepSeek).
Install
# after publishing to GitHub (repo tagged with the `dsh-plugin` topic)
dsh plugin --profile web add github:zhengjy01/dsh-cubox
# local development
dsh plugin --profile web add link:/path/to/dsh-cubox
Restart dsh web. The plugin ships pre-built — lib/index.js is plain ESM.
Configure
Open Cubox preferences → Extensions & Automation → API Extension → enable it and copy your unique link (e.g.
https://cubox.pro/c/api/save/abcd12345).Give the link to the plugin — either in the settings panel (Settings → Cubox), or just ask the agent:
帮我配置 Cubox,API 链接是 https://cubox.pro/c/api/save/abcd12345The agent calls
cubox_configto persist it.cubox.prois the default server;cubox.ccis the international instance (auto-detected from the link).(Optional) Set the export dir + AI brief: choose the local folder, paste an OpenAI-compatible API key (DeepSeek by default), and edit the prompt template.
{collection}is replaced with today's collection list.
Then:
同步一下今天的收藏 → cubox_sync (days=1)
查一下收藏里关于 LLM 的文章 → cubox_cards (keyword="LLM")
The API link and the LLM key are credentials — anyone holding them can read (and modify) your Cubox data or spend your LLM quota. They are stored in
~/.dsh/dsh-cubox.json(mode 0600). To rotate, refresh in the Cubox API-extension page, then re-configure.
Tools
| Tool | Purpose |
|---|---|
cubox_status |
Connection & cache status |
cubox_config |
Set / clear apiLink, server, token, syncMinutes, outputDir, exportCards, llmBaseUrl, llmApiKey, llmModel, llmPrompt |
cubox_sync |
Pull the last N days (default today) into the local cache + export markdown + AI brief |
cubox_cards |
Query the collection (keyword / days / annotated / starred / read) |
Notes
- The public Cubox "open API" only accepts saves; reads go through the internal
/c/api/cli/*family (same endpoints as the official cubox-cli). - The sync cache is a plain JSON snapshot (cards + today's annotations, deduped by id, newest first) — agents can answer "what did I save" without extra round trips.
- License: MIT
Operate deliberately
Install and manage
Prerequisites and target Profile
Target: Web Profile
Delivery: Dsh Bundle Git — zhengjy01/dsh-cubox#84365663d52aea6251865bf18c5584ceae7ab6d7。
Verify, update, and remove
Show lifecycle commands
dsh plugin --profile web listCompatibility and access
DeepSeek Harness web bundle; Node 22.19+ or 24+: Not declared in supplied evidence。
Review compatibility evidence ↗
Risk facts
Stores a Cubox API-extension link and optional LLM API key locally; either credential may permit Cubox changes or consume LLM quota.
Evidence ↗Reads Cubox data through internal /c/api/cli endpoints and can call an OpenAI-compatible LLM service for daily briefs.
Evidence ↗Evidence and editorial reviewManifest, Bundle patch, distribution and freshness
Immutable evidence
Review status and source activity
Use the Git bundle installation path documented by the project. Treat the Cubox link and LLM key as sensitive credentials.
AI reviewed Sep 11, 2026, 2:11 PM UTC。GitHub facts last checked Sep 11, 2026, 2:11 PM UTC。
No material source change has been recorded since this evidence baseline.