AI Agent Hub
Back to plugins
🧩

dsh-time-travel

admin-security Updated 2026.08.25

Run the following command in DeepSeek Harness:

dsh plugin install helibeiqi/dsh-time-travel

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install helibeiqi/dsh-time-travel in the DeepSeek Harness terminal; the source is at https://github.com/helibeiqi/dsh-time-travel . Restart dsh web (port 3080) after installation to activate the plugin.

About this plugin

AI models mutate files, run shell commands, and fire off HTTP requests across multi-turn tool calls. When something goes wrong in a later turn, the workspace is often left in a state that is painful to undo. dsh-time-travel exists to make every side effect reversible: it builds a snapshot-before, execute, compensate-afters pipeline for fs.write, fs.delete, bash, and http.request so no file mutation is irrecoverable.

The core API is the ctx.timeTravel service. rewindTo(turnId) walks compensation steps in reverse order to restore the workspace to the state just before a given turn; rewindAll rolls back every recorded side effect in one call. For operations that cannot be auto-compensated (arbitrary bash commands, HTTP requests), the plugin does not guess. It marks them as manual and defers to a human. In parallel, the built-in audit rule engine watches for dangerous shell invocations, sensitive-path writes, and non-compensable writes, appending one JSON line to audit-log.jsonl each time so every high-risk action is on record.

It is aimed at developers running DSH multi-step agent workflows who need file-level reversibility and an audit trail. It complements session-level replay or turn-rewind plugins: where those handle conversation history, dsh-time-travel handles the filesystem itself.

Use Cases

  • Roll back the workspace to a prior turn after tool calls corrupted files
  • Detect dangerous shell invocations and sensitive-path writes to ensure compliance
  • Automatically deregister all hooks on plugin unload without polluting the tool pipeline

Best For

  • Developers running DSH multi-step agent workflows
  • Teams requiring file-level reversibility and audit trails
  • Engineers focused on AI tool side-effect safety and compliance