agi-fans/dsh-better ↗★ 1

@agi-fans/dsh-better

Deepen prompts before sending and check subagent briefs in DeepSeek Harness (adapted from pi-better).

パッケージ
@agi-fans/dsh-better
互換性
未検証
バージョン
0.1.2
ライセンス
MIT
最終更新
2026/11/14

インストール

$npx -p @deepseek-ai/dsh dsh plugin --profile web add @agi-fans/dsh-better

ドキュメント

README 全文を読む ↗

Configuration

Configure @agi-fans/dsh-better via Settings → Prompt Deepener, Plugins → Prompt Deepener, the /better-model slash command, or cordis.patch.yml defaults:

- id: dsh-better
  name: '@agi-fans/dsh-better'
  config:
    # Inherit the active session's model when null (reuses KV prompt cache), or pin e.g. "deepseek-official/deepseek-reasoner"
    model: null
    # Equivalent to pi-better's `thinking`: defaults to fast `"low"` reasoning (preserving KV cache) when null, or set `"inherit"` / `"off"` / `"low"` / `"medium"` / `"high"` / `"max"`
    reasoningEffort: null
    # Optional override specifically for /better subagent brief checking (falls back to `model`)
    briefModel: null
    # Optional reasoning effort override for /better subagent brief checking (falls back to `reasoningEffort`)
    briefReasoningEffort: null
    # Set false to disable automatic /better brief checking on subagent activations
    checkSubagents: true

UI and /better-model changes are stored atomically in /dsh-better.settings.json and take effect immediately.