dsh-session-rerun
Run the following command in DeepSeek Harness:
dsh plugin install liuyangdongdong/dsh-session-rerun
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install liuyangdongdong/dsh-session-rerun in your DeepSeek Harness terminal to install the plugin from npm (source repository: https://github.com/liuyangdongdong/dsh-session-rerun).
About this plugin
After a multi-step Agent run finishes, the durable log holds every input and output, yet there is no intuitive surface to review or replay individual steps. dsh-session-rerun adds exactly that to DeepSeek Harness Web: it projects the session timeline grouped by Turn and Step, showing model replies, tool calls, Skills invocations, failure points, and elapsed time at a glance, with compact file-name badges on the right marking referenced scripts and Markdown documents.
Once you have located a step worth revisiting, you can trigger a replay in one click. The plugin spawns a brand-new session in the same workspace, feeds the new Agent the durable history preceding the chosen step plus that step's logged user input, and leaves the source session untouched and read-only. For Subagent steps it goes a step further: a new main session and a new child session are created in parallel, the child replays the selected step while the main session reconstructs only the delegation call, waits for the child result, and then continues from there, faithfully mirroring a multi-Agent hand-off. Nested Subagent replay is intentionally not supported, and a step must have a complete step/end event pair before it can be replayed.
This plugin is built for developers who iterate on multi-step Agent workflows in DeepSeek Harness Web or Desktop and need to reproduce a single step repeatedly to tune tool-call behaviour or prompt wording. It ships as a standalone npm package and, once installed, exposes an independent timeline panel with replay controls in the browser—no modifications to the agent-loop source are required.
Use Cases
- Replay a specific step in a new session to verify a fix for a tool-call deviation
- Review a multi-step session timeline to quickly locate failure points and time outliers
- Isolate and reproduce a single Subagent delegation step to debug multi-Agent hand-offs
Best For
- Developers debugging multi-step Agent workflows in DeepSeek Harness Web or Desktop
- Engineers who need to repeatedly reproduce a single step to tune prompts or tool configurations
- Developers reviewing Agent session logs to locate tool-call and Skills invocation anomalies
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.