AI Agent Hub
Back to plugins
dsh-council preview

dsh-council

Workflow Updated 2026.09.16

Run the following command in DeepSeek Harness:

dsh plugin install a1exsun/dsh-council

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

Install the plugin from https://github.com/a1exsun/dsh-council by running dsh plugin install a1exsun/dsh-council inside DeepSeek Harness; it pulls a prebuilt package from npm and requires no compilation.

About this plugin

Dsh-council elevates a single prompt into a structured deliberation: two to eight models answer independently in parallel, one to eight reviewers compare the anonymized answers to surface gaps and produce rankings, and a single arbiter synthesizes everything into a final recommendation annotated with confidence notes, disagreements, and blind spots. The entire pipeline runs inside DeepSeek Harness, and the raw answers and reviews are preserved in DSH child sessions for full traceability.

Each role (answerer, reviewer, arbiter) can be assigned a different model from your configured DSH providers, or the same model can fill multiple roles. A single run involves four to seventeen model participants, each of which may make multiple requests and use web tools. Output is bilingual, following DSH's language setting, and includes consensus points, divergences, blind spots, model identities, and average rankings so you can quickly gauge how solid the conclusion truly is.

This plugin is aimed at engineers and architects who already run DeepSeek Harness and need cross-validated reasoning for high-stakes decisions such as architecture trade-offs, crash-recovery design, or technology comparisons. When a single model's answer, however confident, is not enough, dsh-council turns a one-shot question into a small model council with minimal setup and no extra compilation.

Screenshots

Use Cases

  • Cross-validate architecture trade-offs with parallel model reasoning and peer review
  • Stress-test crash-recovery and failover designs through multiple independent perspectives
  • Surface blind spots in technology comparisons and get an arbitrated final recommendation

Best For

  • System architects who need cross-validated reasoning for design decisions
  • Advanced engineers already operating within DeepSeek Harness
  • AI product leads making high-stakes technology calls