feeder

星标8

分支13

更新时间2026年7月20日 01:09

Find, deduplicate, review, and normalize factor research sources into alpha_resource.md for AgentMatrix factor_lab.

安装

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

在 Manus 中运行

来源

AgentMatrixLab

AgentMatrixLab/agentmatrix-research

打开 GitHub 仓库查看创作者相关仓库

下载

在 Manus 中运行

AgentMatrix Feeder Skill

Use this skill when an agent needs to turn a paper, broker report, local PDF, web page, or structured note into a testable alpha_resource.md.

Rules

Do not claim alpha validity. This skill only prepares a source for factor research.
Reject vague strategy commentary, performance-only summaries, and black-box descriptions that cannot become a factor formula or data source.
Preserve source provenance, formula assumptions, required fields, frequency, universe, rebalance rule, horizon, and expected direction.
Do not bypass paywalls, CAPTCHA, login restrictions, or vendor terms.
Do not ask users to paste credentials, cookies, tokens, private keys, or passwords into chat.

Required Output

Write one alpha_resource.md containing:

source title, author/institution, date, URL or local path
novelty and duplicate notes
factor family and formula
required fields and field definitions
frequency, universe, horizon, rebalance rule, and neutralization assumptions
implementation uncertainties and stop conditions
recommended factor_lab command for the next stage

Handoff

Pass alpha_resource.md to the factor research skill. If the source is not implementable, stop with a short handoff explaining the missing formula, data fields, or point-in-time assumptions.

同仓库更多 Skills

同仓库

quant-api-v2

AgentMatrixLab/agentmatrix-research

通过 HTTP API 查询/拉取 A 股量化数据. 25 张 CH 表 (33 因子/月频/日 K/分钟 K/财务/行业/市值/停牌/ETF/交易日历/行业指数/交易所指数/指数权重/行业成分/交易所成分/L1 快照), 4 个本地 parquet (market_cap/pe/pb/ps), 33 个月度因子 (41.9 万行, 76 月), 加 RQData 异步拉取 (alpha101 等因子). 支持 JSON / NDJSON 流式 / Parquet 二进制 / CSV 四种下载格式. 触发场景: 用户要求查 ROE/PE/PB/动量等因子, 拉日 K 线, 读 parquet 文件, 探索数据源, 查 API 端点, 因子 IC 评估, 触发 RQData 拉新因子数据 (alpha001-101 等), 下载大数据表为 Parquet.

2026-07-268

agentmatrix-alpha-pipeline

AgentMatrixLab/agentmatrix-research

Run the AgentMatrix alpha pipeline from alpha_resource.md through internal validation, strategy packaging, external simulation, and live-readiness gates.

2026-07-208

factor-research-proof-pipeline

AgentMatrixLab/agentmatrix-research

Standardizes factor reproduction, truth validation, and proof export in factor_lab. Invoke when building or reviewing Alpha101, Alpha191, Alpha158, Barra, or paper-derived factor workflows.

2026-07-208

alpha101-factor-research

AgentMatrixLab/agentmatrix-research

Builds and validates Alpha101 factors in factor_lab. Invoke when reproducing Alpha101, exporting proof artifacts, or preparing intern/agent research runs.

2026-06-238

name	feeder
description	Find, deduplicate, review, and normalize factor research sources into alpha_resource.md for AgentMatrix factor_lab.

AgentMatrix Feeder Skill

Use this skill when an agent needs to turn a paper, broker report, local PDF, web page, or structured note into a testable alpha_resource.md.

Rules

Do not claim alpha validity. This skill only prepares a source for factor research.
Reject vague strategy commentary, performance-only summaries, and black-box descriptions that cannot become a factor formula or data source.
Preserve source provenance, formula assumptions, required fields, frequency, universe, rebalance rule, horizon, and expected direction.
Do not bypass paywalls, CAPTCHA, login restrictions, or vendor terms.
Do not ask users to paste credentials, cookies, tokens, private keys, or passwords into chat.

Required Output

Write one alpha_resource.md containing:

source title, author/institution, date, URL or local path
novelty and duplicate notes
factor family and formula
required fields and field definitions
frequency, universe, horizon, rebalance rule, and neutralization assumptions
implementation uncertainties and stop conditions
recommended factor_lab command for the next stage

Handoff

Pass alpha_resource.md to the factor research skill. If the source is not implementable, stop with a short handoff explaining the missing formula, data fields, or point-in-time assumptions.