AI Agent Hub
Back to plugins
🖥️

dsh-observability

Client Updated 2026.08.15

Run the following command in DeepSeek Harness:

dsh plugin install jeremy9682/dsh-observability

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

Run dsh plugin install jeremy9682/dsh-observability in your DeepSeek Harness terminal to install the plugin; the full source is available at https://github.com/jeremy9682/dsh-observability

About this plugin

When orchestrating multiple external agents inside DeepSeek Harness, the hardest thing to track is rarely the output itself but which agent touched which session and in what state it was left. dsh-observability closes that gap by fingerprinting product session roots such as ~/.codex/sessions and ~/.cursor/chats before and after each tracked tool call, then writing a lightweight attribution envelope (sessionRef, sessionStatus) to a journal file. Pointers only — no prompt or transcript content is ever persisted, keeping the trail minimal yet fully auditable.

The second piece, dsh-tool-cursor, exposes a cursor_delegate tool that delegates a task to the official cursor-agent CLI in headless mode. It is read-only by default; in edit mode it proposes changes but never applies them, keeping every external agent inside a controlled boundary. A standalone script, bin/dsh-run-attributed.mjs, is also included. It wraps any CLI with the same fingerprint, envelope, and three-tier timeout (first-event / idle / total) protection without requiring a DSH environment at all.

Who it is for: engineering teams running multi-agent workflows who need to audit attribution, debug session states, or safely invoke the Cursor headless agent from within Harness. If you want a lightweight, traceable observability link with zero extra infrastructure, this plugin fills exactly that gap.

Use Cases

  • Trace session attribution and state for each external agent in multi-agent workflows
  • Safely invoke the Cursor CLI in headless mode with read-only or propose-only controls
  • Attach fingerprint, envelope, and three-tier timeout protection to any CLI without DSH

Best For

  • Engineering teams running multi-agent workflows and needing attribution audits
  • Developers who need to safely invoke external agent CLIs inside Harness
  • Architects who want a lightweight observability link with zero extra infrastructure