rollingSirius/equity-research-skill

Possibly the deepest AI equity-research skill: nine-chapter single-stock deep dives and earnings deep-dives, with scripted DCF/EPV/EVA and reproducible valuation. Covers US, HK and A-shares. Docs in EN and ZH.

What it solves

This project provides a structured framework for AI agents to perform professional-grade equity research. It moves beyond simple stock summaries to produce deep, institutional-quality reports that are fact-traceable, valuation-reproducible, and conclusion-auditable, specifically focusing on the "expectations gap" between market pricing and independent analysis.

How it works

The system operates as a "skill" (a set of instructions and scripts) that an AI agent follows. It employs a rigorous six-step workflow and utilizes specific tools for different research needs:

  • Research Modes: It offers a "Full deep-dive" for new coverage and a "Deep earnings mode" for analyzing quarterly or annual results.
  • Valuation Engine: It uses scripts/dcf.py to execute multiple valuation methods (DCF, Reverse DCF, EPV, EVA, SOTP, and Monte Carlo) to ensure results are not guessed but calculated.
  • Quality Control: A dedicated checker (scripts/check_research_output.py) performs forensic accounting reviews (e.g., Beneish M-Score, accrual quality) to assign an earnings-credibility grade (A-D).
  • Industry Specialization: It routes companies to one of 20 industry-specific appendices to ensure the correct KPIs and valuation models are used (e.g.,ARR for SaaS, NIM for Banks).
  • Source Discipline: It enforces a strict hierarchy of data sources, prioritizing regulatory filings and official documents over second-hand media reports.

Who it’s for

Serious investors, buy-side analysts, and anyone requiring audited, deep-dive investment memos and long-term company coverage rather than quick, one-line stock tips.

Highlights

  • Expectations-Gap Focus: Centers reports around what the market has already priced in versus the analyst's view.
  • Forensic Accounting: Automatically checks for revenue-recognition red flags and earnings quality before valuation.
  • Multi-Method Valuation: Requires cross-checking at least three different valuation methods to avoid bias.
  • Industry-Aware: Includes specialized frameworks for 20 different sectors to ensure relevant metrics are tracked.
  • Market Coverage: Supports US, HK, and A-share markets, including specific handling for China ADR/VIE structures.

Related

  • Project
  • Project
  • Project
  • Project
  • Project