agentsclimarketplace
Sign in

Quant oral to code

Skill 2033121/quant-oral-to-code/skills/quant-oral-to-code

把口述或抽象策略想法转换成可回测代码骨架、回测入口和结果报告，并在生成前做 DSL 提炼与风险护栏审查。Invoke when 用户希望把量化思路直接写成策略代码。From its SKILL.md

Install
npx -y skills add 2033121/quant-oral-to-code --skill quant-oral-to-code

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
  • runs commandsInstructs the agent to run 1 command, including `python "skills/quant-oral-to-code/scripts/validate_quant_oral_to_code.py"`.

SKILL.md

2.0 KB, 619 tokens by cl100k_base, as published. Nobody here has run it

quant-oral-to-code

这个 skill 的目标是把自然语言策略描述整理成一条清晰链路：

  1. intake 收集策略口述、数据条件和输出预期。
  2. dsl 把口述压缩成结构化策略规格。
  3. guardrails 审查前视、过拟合、样本外缺失和乐观成交假设。
  4. codegen 生成策略代码骨架与回测入口。
  5. report 输出交付说明、限制和下一步建议。

当前阶段只提供最小可运行骨架，重点是把模块边界和验证入口搭起来，供后续任务继续填充脚本、模板、schema 和例子。

固定口径

  • 标准数据归一化目标固定为 DuckDB；后续导入 CSV/Parquet/JSON/SQLite 时，也要先落到 DuckDB 再进入统一回测路径。
  • claim level 的基础枚举至少包含 demo_only、portable_backtest、research_grade_local。
  • 缺少真实数据、字段映射或关键定义时，必须进入 cutoff / DATA_REQUIRED 路径，而不是伪装成正式回测完成。

使用顺序

默认按以下顺序读取模块：

  • modules/preflight.md
  • modules/intake.md
  • modules/dsl.md
  • modules/guardrails.md
  • modules/data.md
  • modules/codegen.md
  • modules/onboarding.md
  • modules/report.md

当前交付约束

  • 只承诺生成骨架，不承诺已经具备完整回测能力。
  • 没有真实数据时，必须明确说明不能伪装成研究级回测结论。
  • 输出中要显式标记假设、缺口和需要用户补充的数据。
  • 当前默认链路已经接入 Task 2 的权威模块：preflight、DuckDB 数据契约、claim gate 和 onboarding 文案。

验证

结构验证脚本：

python "skills/quant-oral-to-code/scripts/validate_quant_oral_to_code.py"

What ships with it: 84 files

211.6 KB alongside SKILL.md, 43 of them executable

examples/

modules/

references/

schemas/

scripts/

44 more files not listed here. See all 84 in the repository.

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.