by cpj-dev
Bridge Deepseek-harness into Claude Code for review, critique, delegation, and session import.
# Add to your Claude Code skills
git clone https://github.com/cpj-dev/dsh-plugin-ccGuides for using ai agents skills like dsh-plugin-cc.
Last scanned: 8/25/2026
{
"issues": [],
"status": "PASSED",
"scannedAt": "2026-08-25T04:38:10.652Z",
"npmAuditRan": true,
"pipAuditRan": true,
"promptInjectionRan": true
}dsh-plugin-cc is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by cpj-dev. Bridge Deepseek-harness into Claude Code for review, critique, delegation, and session import. It has 107 GitHub stars.
Yes. dsh-plugin-cc passed SkillsLLM's automated security scan — a dependency vulnerability audit plus prompt-injection heuristics — with no high-severity issues. You can read the full report in the Security Report section on this page.
Clone the repository with "git clone https://github.com/cpj-dev/dsh-plugin-cc" and add it to your Claude Code skills directory (see the Installation section above).
dsh-plugin-cc is primarily written in JavaScript. It is open-source under cpj-dev on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh dsh-plugin-cc against similar tools.
No comments yet. Be the first to share your thoughts!
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM is an informational directory and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
Claude Code plugin that runs DeepSeek Harness (dsh) from slash commands: review, critique, one-shot tasks, and resumable multi-turn sessions.
Pin: @deepseek-ai/dsh@0.1.1-rc.2. After upgrading the plugin, rerun /dsh:setup. Re-verify docs/dsh-compat.md on every dsh upgrade.
Default is standard. minimal and anchored-standard are switches — they are not the default.
| Mode | Tools the model sees |
|---|---|
standard (default) |
Full dsh-base catalog from request #1 (search, skills, subagents, …). No overlay. |
minimal |
bash + str_replace_editor for the whole run. Extra tools cannot appear later. |
anchored-standard |
Those two tools first. After this session records a tool call or an assistant reply, the next assemble restores the full catalog. |
Switch:
/dsh:run --mode minimal … or --mode anchored-standardDSH_CC_MODE=minimal/dsh:setup --mode minimalA broker (--session / --resume / /dsh:import) keeps the mode it started with. Mismatch → /dsh:stop --broker.
Needs Node >= 20 and a DEEPSEEK_API_KEY. /dsh:setup also needs Node >= 22.19, npm, and pnpm (corepack enable).
/plugin marketplace add cpj-dev/dsh-plugin-cc
/plugin install dsh@deepseek-dsh
/dsh:setup
/dsh:review
Already have a built deepseek-harness checkout? /dsh:setup --harness <path>. Already have a dsh binary? set DSH_BINARY. Uninstall: remove the plugin, the plugin data dir, and ~/.dsh/profiles/cc.
| Command | What it does |
|---|---|
/dsh:check |
Readiness probe |
/dsh:setup |
Install pinned npm CLI (or --harness) and the multi-turn cc profile |
/dsh:review [focus] |
Read-only review of local changes |
/dsh:critique [focus] |
Structured adversarial critique |
/dsh:run <task> |
Task (--write, --session, --resume, --mode, --background) |
/dsh:delegate <task> |
Background handoff via the dsh-delegate subagent |
/dsh:import |
Weak-import this conversation into a resumable dsh session |
/dsh:runs / /dsh:show |
List runs / replay a stored result |
/dsh:stop / --broker |
Kill a run / the shared broker |
Flags and failure modes: docs/commands.md. Recovery: docs/troubleshooting.md.
Read in this order; stop when you can act.
Entry: plugins/dsh/scripts/dsh-bridge.mjs (one subcommand per capability; stdout is user-facing). DSH argv and overlays: plugins/dsh/scripts/lib/dsh.mjs. Broker: plugins/dsh/scripts/dsh-broker.mjs — started by the bridge; never start it by hand. Tests: npm test (fake dsh, no API key).
Layout: .claude-plugin/marketplace.json · plugins/dsh/commands/*.md · plugins/dsh/scripts/ · docs/. Index: docs/README.md. Chinese user docs: docs/zh-CN/README.md.
--write or not).--session / --resume / /dsh:import record session ids, and those live only as long as the broker process./dsh:import is a compressed text digest, not native history replay. Non-text Claude blocks (including images) are dropped.$ARGUMENTS is text; this plugin never forwards pasted images to DSH. --model deepseek-v4-flash-vision-exp only changes the model id. There is no --image flag. Native DSH vision belongs in dsh web / TUI. In standard, DSH may read_image a repo file — that is not a product path for Claude chat images.Third-party copyrights, licenses, and design provenance live in NOTICE. In short:
--mode anchored-standard: assemble-filter / promotion protocol reimplemented from xiaobright/dsh-anchored-standard (MIT). Not a copy of that Web preset. First-request trigger evidence: xiaobright/modeltest (research citation; no LICENSE file at citation time).This project is not affiliated with or endorsed by those authors or organizations.
MIT — see LICENSE. Redistributors must keep LICENSE and NOTICE.