simon300000/dsh-auto
40Last commit Aug 14, 2026
dsh-auto DSH plugin
dsh-auto adds an Auto Approve permission preset to the DeepSeek Harness Web UI. When a session selects this preset, each approval request is reviewed by a fresh, restricted child agent that can only perform read-only investigations. The plugin enforces a fail-closed policy and limits consecutive denials.
How to install the dsh-auto DSH plugin
dsh plugin --profile web add github:simon300000/dsh-autoCopying does not run this command. Review the repository and version before installing the dsh-auto DSH plugin.
dsh-auto DSH plugin data source
dsh-auto DSH plugin snapshot date: Aug 16, 2026
discovered
What the dsh-auto DSH plugin can do
- Provides an Auto Approve permission preset in the DSH Web UI session permissions selector and default settings.
- Spawns a restricted child agent (Reviewer) for each approval request with a read-only sandbox and limited tool set (read, glob, grep, structured output).
- Enforces a fail-closed security model: denies on timeout, tool failure, invalid output, or missing action data.
- Limits consecutive denials: three denials in the same parent turn interrupt that turn.
- Allows configuration of reviewer provider, model, reasoning effort, timeout, investigation steps, and token limits via YAML.
Where the dsh-auto DSH plugin fits
- Automating approval of low-risk read-only actions (e.g., reading files, listing directories) in a DSH session.
- Reducing manual approval overhead for developers who trust the automated reviewer to handle routine requests.
- Enforcing a security policy where high-risk actions require explicit human authorization after automated review.
- Testing or experimenting with automated approval workflows in a controlled DSH environment.
Who the dsh-auto DSH plugin is for
- DSH Web UI users who want to automate repetitive approval decisions.
- Developers and security engineers managing DSH approval policies and seeking an additional permission preset.
dsh-auto DSH plugin limitations
- Currently supports only the DSH Web UI; not available in other interfaces.
- Requires a configured reviewer provider and model (defaults to deepseek-official/deepseek-v4-flash).
- Reviewer decisions depend on the model's quality and may incorrectly allow or deny actions.
- The plugin has configurable timeouts and token limits that may block complex or long-running approvals.
- Only available as a GitHub plugin; installation requires the dsh command-line tool.
dsh-auto DSH plugin: from the repository README
Quoted from the simon300000/dsh-auto README, the upstream source of the dsh-auto DSH plugin. Copyright remains with the original authors.
English | [中文](README.zh.md) `dsh-auto` adds an `Auto Approve` permission preset to the DeepSeek Harness Web UI. Each action that requires approval is reviewed by a fresh, restricted DSH child Agent before the plugin allows or denies it. The current release supports the Web UI only. ## Screenshots Select the `Auto Approve` permission preset:  The Reviewer allows a bounded read-only action:  The Reviewer denies a high-risk action without sufficient user authorization:  ## How it works - The plugin handles `approval/request` only when the session selects `Auto Approve`. Other permission presets continue through DSH's existing approval chain. - Each approval starts one `spawn` Reviewer session. DSH's own agent loop handles any bounded `read`, `glob`, or `grep` investigation and captures the final structured result; the plugin does not implement a separate model/tool loop. - The child is created with a read-only sandbox and `approval/policy
Read the full READMERepository license: MIT
dsh-auto DSH plugin questions
How do I install dsh-auto?
Install dsh-auto from the GitHub repository using the command: `dsh plugin --profile web add github:simon300000/dsh-auto`. After installation, restart the DSH Web UI. Then select 'Auto Approve' in the session Permissions selector or set it as the default permission preset in General Settings.
How does the Auto Approve plugin work?
When a session uses the Auto Approve preset, every approval request triggers a new child agent (Reviewer) with a read-only sandbox. The Reviewer can only use read, glob, grep, and a structured output tool. It evaluates the action, approval reason, permissions, and context, then returns an outcome (allow/deny). The plugin fails closed on errors and limits consecutive denials to three per turn.
What configuration options are available?
You can customize the reviewer provider and model, reasoning effort, timeout (default 90s), max investigation steps, max consecutive denials, and various token limits. All settings are defined in the plugin's YAML configuration. If you omit reviewerProvider and reviewerModel, the Reviewer uses the parent session's current provider and model.
Why did the Auto Approve plugin deny my action?
The Reviewer denies actions that it considers high-risk or lacking sufficient user authorization. The plugin also denies on critical risk, and on high risk without at least medium user authorization. Technical failures (timeout, tool failure, invalid output) also result in a denial. Review the Reviewer's rationale (if provided) or check the console logs for details.
Can I customize the Auto Approve security policy?
Yes. The policy templates are located in `prompts/policy-template.md` and `prompts/policy.md` inside the plugin directory. You can modify these files to adjust the Reviewer's persona and additional security rules. After changing the policy, restart DSH for the changes to take effect.