alpha-research

Alpha Research Skills

A portable, Markdown-first skill collection for turning vague market ideas into structured, source-grounded, red-teamable investment research reports.

Alpha Research Skills is a source repository for reusable agent skills. It is not a trading bot, financial advisor, brokerage integration, web app, or CLI.

Install

Pi local verification

For a quick Pi session from this repository:

cd /Users/changshun/codes/github/alpha-research
pi -e . --name alpha-research-verify

For a global Pi install from a local checkout:

pi install /Users/changshun/codes/github/alpha-research
pi --name alpha-research-verify

For a Pi install from GitHub, use an explicit GitHub source so it is not resolved to another Git host:

pi install git:github.com/sogud/alpha-research

In Pi, skills are invoked with the skill: prefix:

/skill:alpha-research

If Pi is already running after an install, reload resources:

/reload

Open Plugin CLI

Common commands:

# Preview
npx plugins discover sogud/alpha-research

# Auto-detect supported targets
npx plugins add sogud/alpha-research

# Install to one target explicitly
npx plugins add sogud/alpha-research --target claude-code
npx plugins add sogud/alpha-research --target cursor
npx plugins add sogud/alpha-research --target codex

# List detected targets
npx plugins targets

If the repository is private, use an SSH URL instead:

npx plugins add [email protected]:sogud/alpha-research.git

Some plugin hosts namespace installed skills, for example /alpha-research:alpha-research. If you want the un-namespaced /alpha-research skill entry, use the Agent Skills installer below.

Claude Code plugin marketplace

If you use Claude Code v2.1.142+ with plugin marketplace support:

/plugin marketplace add sogud/alpha-research
/plugin install alpha-research@alpha-research

This is optional Claude Code metadata. The canonical skill source remains skills/.

Agent Skills installer

Use explicit GitHub URLs. Avoid wildcard agent targets; choose the runtime you actually use.

# Claude Code
npx skills add https://github.com/sogud/alpha-research -a claude-code -g -y

# Cursor
npx skills add https://github.com/sogud/alpha-research -a cursor -g -y

# Codex
npx skills add https://github.com/sogud/alpha-research -a codex -g -y

For project-local installs, omit -g if your installer supports it:

npx skills add https://github.com/sogud/alpha-research -a claude-code

Manual fallback

git clone https://github.com/sogud/alpha-research.git
cd alpha-research
export AGENT_SKILLS_DIR="$HOME/.agents/skills"
mkdir -p "$AGENT_SKILLS_DIR"
cp -R skills/* "$AGENT_SKILLS_DIR"/

Change AGENT_SKILLS_DIR to your agent's actual skill directory.

Quick start

After installation, start with one skill:

/alpha-research

English prompt:

Use /alpha-research to research AI optical networking, gold, oil, or another commodity-linked theme.
Run the Alpha Research workflow.
Use strict evidence standards.
You may recommend candidate companies only with data, logic, citations, risks, and failure criteria.
Do not provide direct trade signals, target prices, position sizing, or return guarantees.
Separate Fact, Inference, Assumption, and Rumor.
Produce the final report as `report.html`; Markdown artifacts are only intermediate work products. The default report style is Dense Calm + ECharts: high information density, calm visual surface, compact tables, plain-language explanations, and ECharts SVG figures.

中文 prompt:

使用 /alpha-research 研究 AI 光通信、黄金、原油或其他大宗商品相关主题。
直接运行 Alpha Research 完整研究流程。
使用严格证据标准。
可以推荐候选公司，但必须有数据、逻辑、引用、风险和证伪条件。
不要输出直接交易信号、目标价、仓位建议或收益承诺。
必须区分 Fact、Inference、Assumption、Rumor。
最终输出 Dense Calm + ECharts 风格的单文件 HTML 报告；过程产物可以是 Markdown。

What it does

Alpha Research follows a Serenity-style chokepoint research process:

Setup → Orchestrate → Clarify → Profile → Source → Gate → Trace → Audit → Red-team → Synthesis + Quality Gate → Report

flowchart TD
  A[Setup]
  B[Alpha Research]
  C[Thesis Grill]
  D[Source Map]
  E[Chain Trace]
  F[Thesis Audit]
  G[Report Quality]
  H[Final Report]

  A --> B
  B --> C
  C --> D
  D --> E
  E --> F
  F --> G
  G --> H

Alpha Research has a hard source gate: if required primary/data sources, financial baselines, narrative-origin capture, Source Registry entries, citation ids, clickable original-source URLs, or previewable stable locators are missing, the workflow stops before the final report and returns a Source Acquisition Plan.

Workflow

/thesis-grill → /source-map → /chain-trace → /thesis-audit → /report-quality

Anti-Hype is a stricter research mode for testing popular market narratives, social-media claims, and crowded stories.