DeepSeek Harness plugin

dsh-mimo-vision-hint

DSH plugin: prompt the agent to dispatch image recognition to an opencode-go mimo-v2.5 subagent

Jump to install

Source facts

Repository
Isekai-Mfu/dsh-mimo-vision-hint
Latest update
Aug 13, 2026
Category
Workflow & Automation
GitHub stars
2
Format
plugin
Catalog evidence
Upstream dsh.bundle evidence
Evidence path
package.json#dsh.bundle
Checked against
0.1.0-rc.8
Upstream check date
2026-08-20

This evidence comes from the upstream catalog. This site has not installed, run, or security-reviewed the plugin.

Install

Start with a prompt that asks an agent to review the GitHub repository and source. Switch to the command if you want to install it yourself.

Copy this prompt into DSH, Codex, or another agent and ask it to review the GitHub repository and source first.

Do not install or run any commands yet. Read this plugin's GitHub repository, README, and relevant source code. Then answer the questions below clearly and directly so I can decide whether it fits my needs:

1. What is this plugin, and what problem does it solve?
2. Who is it for, and what are its typical use cases?
3. How is it used after installation? Include one minimal example.
4. What known limitations or privacy, security, compatibility, or maintenance risks does it have?
5. Give a clear recommendation: recommend, conditionally recommend, or do not recommend, with reasons.

Distinguish statements documented by the repository, inferences from source code, and unknowns. If evidence is insufficient, say so explicitly. Do not guess or simply repeat the README.

GitHub: https://github.com/Isekai-Mfu/dsh-mimo-vision-hint
Plugin: dsh-mimo-vision-hint
Author: Isekai-Mfu

Check the source files

Read the README and other files from this plugin directory before installing.

File explorer4 files
README.mdSource · read only
README language

dsh-mimo-vision-hint

English | 中文

A zero-dependency hint plugin for DeepSeek Harness (DSH). It registers a section in the model's system prompt instructing the agent to dispatch image-recognition tasks to an opencode-go mimo-v2.5 subagent (a single-agent workflow with provider/model overrides) instead of reading images itself via read_image; if the workflow call fails, the agent falls back to read_image.

This plugin only injects prompt text: it registers no model tools and provides no cordis services, so mounting it in an agent composition needs no isolate realm.

Prerequisites

  • Your DSH deployment has a provider named opencode-go configured, serving a mimo-v2.5 model — the subagent dispatch targets that provider/model pair.
  • The active agent preset includes the workflow tool (needed to spawn the subagent).

Without these, the plugin does no harm: the injected prompt tells the agent to fall back to read_image directly.

Installation

This package is a DSH profile bundle (its manifest declares dsh.bundle and it ships its own mount layer), so the official plugin command installs and mounts it in one step — no hand-edited YAML required:

dsh plugin --profile web add github:Isekai-Mfu/dsh-mimo-vision-hint

dsh plugin installs the package into the profile's node_modules, records the dependency, and — because this package declares a bundle — automatically appends its layer ([cordis.patch.yml](cordis.patch.yml), a single insert entry) to the profile's dsh.profile.bundles composition stack. The plugin then mounts at the host layer and takes effect for every new session of every agent preset in the deployment.

pnpm must be on your PATH (dsh plugin forwards to it). Uninstalling is equally one command — once the dependency is gone, the bundle layer leaves the composition stack automatically:

dsh plugin --profile web remove dsh-mimo-vision-hint

Manual installation (without pnpm)

1. Place this package somewhere the profile can resolve (e.g. $DSH_HOME/profiles/web/node_modules/dsh-mimo-vision-hint). 2. Add an insert entry to the profile's user patch layer ($DSH_HOME/profiles/web/cordis.patch.yml):

- insert:
    - id: mimo-vision
      name: dsh-mimo-vision-hint

3. Saving takes effect immediately (hot reload); remove the entry to disable.

Notes

No other dependencies, no configuration options. The prompt text is injected into every new session; the vision subagent's provider (opencode-go) and model (mimo-v2.5) are currently hard-coded — edit SECTION in lib/index.js to change them. Presets without the workflow tool (e.g. minimal mode) cannot dispatch the subagent; the prompt tells the model to fall back to read_image in that case.

License

[MIT](LICENSE)