AI Agent Hub
Back to plugins
🧩

dsh-self-control-guard

admin-security Updated 2026.08.17

Run the following command in DeepSeek Harness:

dsh plugin install pandashere/dsh-self-control-guard

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 pandashere/dsh-self-control-guard, with the full source code available at https://github.com/pandashere/dsh-self-control-guard.

About this plugin

During DeepSeek Harness plugin development, agents often accidentally terminate the host process with commands like pkill dsh or kill -9, leading to interrupted sessions, lost work, and no audit trail. The dsh-self-control-guard plugin addresses this by acting as a self-control guard for the host process. Its core capabilities include intercepting high-confidence kill attempts from the bash tool, hard-denying them, and teaching the model to use controlled exit tools such as dsh_self_exit, while logging an audit trail for each interception. The plugin focuses on preventing accidental self-termination without altering the existing command execution flow, and it employs token-confirmed graceful exits or restarts. It is ideal for DSH plugin developers who seek to enhance session stability, avoid unintended interruptions, and maintain operational logs. Note that this is a UX interception layer, not a security boundary, making it suitable for risk mitigation in standard development environments.

Use Cases

  • Agents accidentally executing kill commands during development causing session termination
  • Intercepting high-risk host termination attempts with logging
  • Controlling plugin agent exit behavior for graceful shutdown

Best For

  • DeepSeek Harness plugin developers
  • Test engineers requiring stable session environments
  • System administrators focused on operational auditing and security