DeepSeek Harness (DSH) adopts a plugin-based architecture, so extending functionality does not require modifying the core code. When building agents or handling specific interaction logic, it may sometimes be necessary to intervene in or correct certain “aggression”-related or specific behavioral patterns. dsh-aggression is the tool provided for this use case.
Overview¶
dsh-aggression is a plugin classified as admin-security and maintained by developer uckkk. It mainly addresses how to detect behavior and provide correction suggestions based on input parameters.
Core Functionality¶
The plugin registers a tool named aggression.
When the tool is called with parameters, it returns correction suggestions for the behavior based on the parameter content.
Installation and Enabling¶
Install it using the official command:
dsh plugin add github:uckkk/dsh-aggression
After installation, the plugin takes effect in the current session, and its registered tool can be called.
Typical Usage¶
In a DSH interactive session, call the registered tool directly. For example, by inputting parameters that contain aggressive or specific behavioral patterns, the plugin returns the corresponding correction suggestions.
Use Cases and Notes¶
- Implementation: The plugin is implemented entirely in Node.js, does not require network access at runtime, and completes logic processing locally.
- Permissions: The plugin runs with the permissions of the current DSH process. Before installation, you should inspect the source code and license to confirm that its system access scope is under control.
Summary¶
dsh-aggression provides the ability to return correction suggestions for specific behavioral parameters, making it suitable for scenarios that require local behavior validation. For more details, see catalog_url or github_url.