dsh-always-require-tools-approval
Run the following command in DeepSeek Harness:
dsh plugin install J0ss077/dsh-always-require-tools-approval
Paste the following prompt into your AI chat to install this plugin:
To install this plugin in DeepSeek Harness, run the command 'dsh plugin install J0ss077/dsh-always-require-tools-approval', with the full source code available at https://github.com/J0ss077/dsh-always-require-tools-approval.
About this plugin
When running agents in DeepSeek Harness, tool executions like bash commands can pose security risks, as the sandbox only blocks file writes but not command execution. This leaves users with limited control over agent actions, potentially leading to unintended or malicious behavior.
The dsh-always-require-tools-approval plugin addresses this by adding an approval gate. When the agent calls a tool on the watchlist (defaulting to bash and pwsh), execution pauses and waits for explicit user approval. Each call requires separate approval, denying if rejected or no approval is given, while other tools remain unaffected. This design ensures that agent tool executions are always under user supervision.
This plugin is ideal for developers, team administrators, or anyone using DeepSeek Harness who wants to enhance agent workflow security. It provides a simple yet powerful way to control agent behavior, preventing unauthorized actions while maintaining workflow flexibility.
Use Cases
- When the agent calls tools like bash, the plugin pauses execution and waits for user approval.
- In team collaboration, ensure agent operations are authorized before running.
- Use in high-security environments to prevent agents from executing sensitive tools without approval.
Best For
- Developers using DeepSeek Harness
- Team administrators who need to monitor agent behavior
- Users who prioritize security in automation workflows
Related Plugins
A method pack that makes AI coding agents plan against your real baseline, prove completion with fresh evidence, and reduce reworks and unsafe changes.
Turns the DeepSeek Harness session into a captain that builds a durable sub-agent team, splits goals into dependency-aware tasks, and coordinates work via direct messages and a live Web UI.
Gives coding agents design judgment, letting Claude Code, Cursor, and 70+ agents generate and iterate high-quality UI, presentations, and graphics right from the terminal.
Run the Pi ecosystem's plugins on DeepSeek Harness, unmodified, via a compatibility layer that implements Pi's public extension ABI on DSH's native services.