hezhongtang/dsh-capability-optimizer ↗★ 0
dsh-capability-optimizer
External-expert consultation for DeepSeek Harness: headlessly invokes the Claude Code CLI with role personas (advisor / reviewer / designer, extensible via config) and returns the subscription model's replies as reference answers for the agent to weigh. · DSH 外部专家咨询：headless 调用 Claude Code CLI，以 advisor / reviewer / designer 等角色人设（可通过配置扩展）获得订阅模型的回复，作为给 agent 权衡的参考答案。
安装
npx -p @deepseek-ai/dsh dsh plugin --profile web add github:hezhongtang/dsh-capability-optimizer说明文档
阅读完整 README ↗Usage
Ask your agent:
"consult the reviewer on this diff before we call it done"
The agent picks the role, packs the material into context, and calls consult_expert. Claude's reply returns as a reference answer with run metadata (session, turns, duration, cost) — advice the agent weighs, not an order it obeys.
| Tool | Read/Write | Purpose |
|---|---|---|
consult_expert | read* | One role, one question, optional material in context, optional model / effort overrides |
consult_panel | read* | Several roles, one question, parallel — all perspectives back together |
consult_roles | read | The live roster (including role-level model/effort) and global defaults |
* Read-only for your workspace; each call spends your Claude subscription quota — the tool descriptions themselves tell the model to batch material instead of machine-gunning calls.