AI Agent Hub
Back to plugins
🖥️

dsh-ultracode

Client Updated 2026.08.14

Run the following command in DeepSeek Harness:

dsh plugin install ParticleLight/dsh-ultracode

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

Run dsh plugin install ParticleLight/dsh-ultracode in your terminal, then restart dsh web to activate the preset; full source repository: https://github.com/ParticleLight/dsh-ultracode

About this plugin

Most agent presets simply crank up the reasoning effort without changing how work actually gets done: half-finished code that skips tests, serial execution of every task, and a final claim of completion with zero evidence. dsh-ultracode defines UltraCode as a complete working mode, not just an effort level. Think at maximum depth before acting; follow a strict plan-execute-verify cycle with tests, diff review, and acceptance criteria that prove the code runs; delegate independent work to subagents or workflows in parallel while you orchestrate and sign off; produce full implementations, not sketches; and close every session by stating exactly what was verified and how.

The preset is built on the DeepSeek Harness code (PTC) foundation, inheriting the full tool catalog, Code Mode SDK presentation, plan mode, compaction, goals, subagents, workflows, and Ralph. It layers the UltraCode working-mode guide into the persona without removing any existing capability; it constrains and strengthens behavior on top of everything already there.

Ideal for developers and teams who demand high-quality, verifiable deliverables from every agent session and want to boost throughput via parallel subagent orchestration. If you are tired of half-baked code output and unproven completion claims, this preset was designed for your workflow.

Use Cases

  • Decomposing and parallelizing complex multi-module tasks with subagents
  • Requiring verifiable complete implementations backed by tests and diff review
  • Expecting sessions to close with explicit evidence rather than mere completion claims

Best For

  • Developers who demand strict quality from agent output
  • Teams using DeepSeek Harness for daily development workflows
  • Tech leads seeking higher throughput via subagent orchestration