dsh-debate
Run the following command in DeepSeek Harness:
dsh plugin install shlouai/dsh-debate
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install shlouai/dsh-debate to install this plugin in DeepSeek Harness; the source repository is https://github.com/shlouai/dsh-debate .
About this plugin
Ask your agent a hard question and you receive one confident voice reasoning in a single direction — it will always surface the supporting arguments because nothing in the loop is obliged to attack them. dsh-debate puts a genuine adversary in the circuit: FOR and AGAINST are separate one-shot subagents that never see each other or the judge, and each is instructed to win. Only after the final round does the main agent break its silence and rule on the full transcript.
Each round is driven by debate_round: the pro side speaks first, then the con side reads it and responds. Debaters run on a deliberately read-only tool set (file read, glob, grep, web search) so they can open real repository files and cite external sources rather than rely on memory. The whole debate renders live as a single collapsible card in the Web UI, with built-in Chinese and English dictionaries, and survives a session reload.
If you use DeepSeek Harness for architecture decisions, technology selection, or solution reviews and want to escape the bias of a single confident narrative, dsh-debate costs a few minutes of interaction in exchange for a complete record where the strongest case against your idea was raised on purpose. It is built for engineers and teams that need an auditable trail behind every go or no-go decision.
Screenshots
Use Cases
- Staging a two-round debate over an architecture choice where both sides cite real repository files
- Generating an audited transcript where the strongest counter-case was raised on purpose before a tech migration
- Running an automated debate to surface the weakest argument chain before a team review
Best For
- Engineers using DeepSeek Harness who need an auditable reasoning chain for technical decisions
- Developers who want to break past single-perspective confidence and get genuine adversarial feedback
- Technical teams that need a complete pro and con argument record before a solution review
Related Plugins
A method pack that makes AI coding agents plan against your real baseline, prove completion with fresh evidence, and reduce reworks and unsafe changes.
Turns the DeepSeek Harness session into a captain that builds a durable sub-agent team, splits goals into dependency-aware tasks, and coordinates work via direct messages and a live Web UI.
Gives coding agents design judgment, letting Claude Code, Cursor, and 70+ agents generate and iterate high-quality UI, presentations, and graphics right from the terminal.
Run the Pi ecosystem's plugins on DeepSeek Harness, unmodified, via a compatibility layer that implements Pi's public extension ABI on DSH's native services.