DeepSeek Harness Plugins · Workflow
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
A hot-pluggable DSH task-board GUI plugin offering a five-column kanban, real agent execution, cron scheduling, and system-prompt injection with zero source changes to DSH.
Deterministic workspace regression testing for coding agents: capture corrections as executable cases, verify in isolated Git worktrees, and minimize failure-inducing components.
Diagnose tool-call integrity defects in DeepSeek Harness sessions and enable non-destructive recovery via safe session forking, with live scanning, offline CLI checks, and Web continuation.
Turn Feishu and local materials into verified, structured docs with conclusions, evidence, and actions - zero credentials, one-command install for DeepSeek Harness.
Turns Markdown into a self-contained HTML slideshow, editable PPTX, and JSON manifest with built-in themes and optional custom masters.
One-click rewrite of vague prompts into structured task briefs with validated reference file lists, running in isolated sessions without polluting main chat.
Brings coding-agent execution trajectories to non-coding work, giving content production a durable, inspectable record of how work actually happened.
Records agent failure patterns and success paths, distills local experience after each session, and provides bounded, temporary context for similar tasks without modifying the model or DSH Core.
Estimate concurrency requirements through precise parameter-based calculation and return detailed results, a pure Node workflow tool with no network dependencies.
A KPI gap breakdown plugin that precisely calculates results from given parameters and returns detailed output, built in pure Node with no network dependency.
A lightweight tool that precisely calculates cleaning frequency based on parameters and returns clear recommendations, built in pure Node with no network dependencies.
Breaks down follower growth goals by precise parameters with a pure Node implementation and no network dependency, returning detailed results.