AI Agent Hub
Back to plugins
⚙️

dsh-pilot

Workflow Updated 2026.08.25

Run the following command in DeepSeek Harness:

dsh plugin install Viger1/dsh-pilot

Paste the following prompt into your AI chat to install this plugin:

Please run `dsh plugin install Viger1/dsh-pilot` in DeepSeek Harness to install this plugin (Source URL: https://github.com/Viger1/dsh-pilot).

About this plugin

In the current dsh ecosystem, many browser plugins only rely on admission checks at the tool entry point, making agents extremely vulnerable to redirects, in-page clicks, and history navigation. Moreover, these plugins often invent their own permission policies instead of respecting the user's session configuration, leading to blurred security boundaries. dsh-pilot solves this problem by enforcing a permission model at the network layer. It reads the dsh session's approval stance directly as the policy basis and uses request interception within the browser context to ensure that every main-frame navigation is strictly verified, regardless of how the navigation starts.

As the core "hands" tool for dsh agents, this plugin provides a complete suite of capabilities from navigation, snapshots, to actions. It parses pages based on Playwright's accessibility tree, allowing agents to precisely control page elements via references to perform complex tasks like clicking, typing, and screenshots. This mechanism not only avoids the tedious maintenance of traditional selectors but also ensures operational consistency and safety.

This plugin is perfect for developers looking to build highly autonomous, secure, and session-policy-aligned dsh workflows. It eliminates the dependency on vision models, allowing agents to naturally handle form flows and web interactions like humans, while strictly adhering to user permissions for a truly "what you see is what you control" experience.

Use Cases

  • Automating complex registration forms
  • Executing web tests that respect security policies
  • Blocking unauthorized navigation before redirects occur

Best For

  • Developers building dsh workflows
  • Users requiring secure browser control
  • Researchers focused on agent autonomy