AI Agent Hub
Back to plugins
⚙️

dsh-plugin-runbook

Workflow Updated 2026.08.26

Run the following command in DeepSeek Harness:

dsh plugin install zhan-tz/dsh-plugin-runbook

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 zhan-tz/dsh-plugin-runbook. The source code is available at https://github.com/zhan-tz/dsh-plugin-runbook.

About this plugin

In research projects using DeepSeek Harness (DSH), scripts, figures, JSON files, and half-finished directories accumulate faster than one can reconstruct the story, making it difficult to track which script generated which output or which run was valid. Runbook addresses this chaos by providing a persistent, Git-backed research surface where the central question remains visible, ensuring that the work of a research project is never lost in a sea of files.

Rather than being a simple file browser or a summary of chat logs, Runbook acts as an intelligent projection layer that integrates existing DSH sessions, Git commits, and project documents into a unified workspace. Its core capability lies in "Attempt cells," where a stable script maintains a single position for reuse instead of creating new date-based folders. Every run generates an execution record and traceable evidence, with recent images and PDFs displayed adjacent to their sources for easy comparison.

This tool is ideal for researchers and AI Agent developers who require iterative experimentation and rigorous verification. It features a "Safe Rerun" workflow that requires a review and confirmation before execution, preventing accidental overwrites. By leveraging Git as the version system and offering features like session lineage tracking and turn-local artifacts, Runbook transforms scattered files into a coherent, professional research workflow that maintains high standards of traceability and efficiency.

Use Cases

  • Rerun scripts and track experimental evidence
  • Manage project code and artifacts with Git
  • Organize scattered scripts and figures

Best For

  • Researchers conducting deep research projects
  • Developers building AI Agents
  • Experimenters needing strict process tracking