dsh-runbook-ops

Team processes for development, testing, release, and operations tend to live scattered across wikis, meeting notes, and tribal knowledge. When a model working inside a Harness session needs to follow a documented procedure, it has no structured source to lean on. dsh-runbook-ops pulls that scattered knowledge into structured YAML runbooks under .dsh/runbooks in the workspace, giving every step a verifiable, reusable, and auditable home.

The plugin exposes three tools: runbook_list enumerates the runbooks available in the current project, runbook_read loads and validates a specific runbook, and runbook_start expands a validated runbook into step-by-step execution guidance for the model. By design it does not execute shell commands directly; command invocation, approval, and logging remain with the tools and governance Harness already provides, keeping a clean boundary between model flexibility and team security policy. Reads are scoped to the current session workspace by default, with a configurable maxBytes cap and an optional allowAbsoluteRoot toggle for broader path access.

If your team already works in DeepSeek Harness and wants to turn SOPs test checklists, release checklists, incident runbooks into structured, auditable guidance the model can follow instead of re-wrapping prompts every time, this plugin is built for exactly that. It suits platform teams, SREs, and any engineering group that benefits from codifying process knowledge into executable assets.