dsh-build-diff
Run the following command in DeepSeek Harness:
dsh plugin install KeLearns/dsh-build-diff
Paste the following prompt into your AI chat to install this plugin:
Download the plugin source from https://github.com/KeLearns/dsh-build-diff, deploy it to ~/.dsh/profiles/node_modules/@kelearns/dsh-build-diff/ with the cordis.patch.yml entry appended per the README, and restart DSH to enable change review.
About this plugin
When you drive an agent workflow inside DSH, every loop cycle leaves a trail of file changes scattered across your workspace. Telling at a glance what the agent touched, how much it changed, and which edits deserve a rollback is painful. dsh-build-diff fills that gap: the moment each agent transition from running to idle completes, the plugin fingerprints the entire workspace, captures content for text files, and organizes every add, modify, and delete under a per-turn record.
The review UI lives in the right sidebar, listing changed files per turn with inline line-level diffs—dual gutters, color-coded +/− lines, collapsed unchanged context. Not happy with what you see? Undo an entire turn in one click, or revert a single file to its exact pre-loop state, including deleting files the agent created and restoring files it removed. A compact review bubble also appears right above the assistant message action buttons, summarizing file count and line delta so you can jump in without opening the sidebar explicitly.
The plugin ships with zero runtime dependencies and does not patch DSH source. Its snapshot engine caches by path, size, and mtime, so unchanged files cost nearly zero IO on subsequent turns. It is built for developers who regularly steer agents to write, refactor, or batch-edit code inside DSH and still want full audit and rollback control over every turn.
Use Cases
- Review all file changes after an agent finishes a refactoring loop and decide whether to roll back
- Trace back through per-turn change history after multiple agent iterations
- Restore a file the agent accidentally deleted to its exact pre-loop state with one click
Best For
- Developers who regularly steer agents to write or refactor code inside DSH
- Tech leads who need audit and rollback control over every agent turn
- Team engineers using DSH workflows who lack a built-in change-audit mechanism
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.