Labs · Agent skill · Claude Code plugin
TopxAI Equity Research
A free, open-source agent skill and Claude Code plugin that runs a research desk: filings-first data from SEC EDGAR, falsifiable calls, a bull/bear debate before every rating, adversarial audits and a ledger that settles every call.
Download the full package (version 1.0.0, Apache-2.0) · Source on GitHub
Why we built it
An AI agent can summarise ten filings in a minute. That is not research. Markets, earnings and news are outcomes; a report is worth paying for when it explains the process behind them, what comes next, and how you will know you were wrong. This skill is the method behind the research we publish, written down so an agent can follow it.
How a report gets made
- Kickoff: pick the topic against other candidates, read the lessons from past calls, fix the price basis, the financial period and the filing cutoff.
- Research lanes: financials, contracts, governance, peers and market, each with a source and an evidence grade for every number.
- Debate: a bull agent and a bear agent argue three points each from the notes alone, then rebut each other once.
- Writing and judgment: a fixed chapter structure, valuation numbers from one model file, and a rating that can be settled.
- Audit: two adversarial rounds, including a skeptic who changes the inputs to see what would flip the rating.
- Publish and watch: every call goes into a ledger, new SEC filings are checked against the conditions that would prove a call wrong, and each call is settled with one lesson.
Install
- Claude Code:
/plugin marketplace add UnbelievableT/topxai-equity-research, then/plugin install topxai-equity-research@topxai. - Codex, Cursor, Gemini CLI, OpenClaw and other Agent Skills tools:
npx skills add UnbelievableT/topxai-equity-research. - Or unzip the download into your skills folder.
The tools only contact sec.gov and data.sec.gov and write files inside your project. Research output is general information, not personalised investment advice.