| name | company-report |
| description | Generate a professional 2-page report on a publicly-traded company as a self-contained HTML website. Given a company name, orchestrates five subagents in order (Generalist, Data Retriever, Data Scientist, Financial Analyst, Report Writer) — pulling the latest 10-K from SEC EDGAR, cleaning the financials, analyzing them, and writing the report. Use when the user asks for a company report, equity write-up, or 10-K-based analysis. Triggers include "write a report on <company>", "company report for <X>", "/company-report". |
Company Report
Produce a professional ~2-page corporate report on a publicly-traded company for
a portfolio manager, delivered as a single self-contained report.html. The work
runs as a five-stage pipeline of subagents that pass data through files in a
per-run working directory.
Input
A company name or ticker (e.g. "Apple", "AAPL", "Coca Cola"). If the user did not
name a company, ask for one before starting.
Setup
- Pick a working ticker/slug for the run (use the obvious ticker if the user
gave one; otherwise a short uppercase slug — the Data Retriever will confirm
the real ticker).
- Create the working directory
output/<TICKER>/ (and output/<TICKER>/raw/).
- This directory path is passed to every stage; agents read prior stages'
outputs from it and write their own outputs into it.
Pipeline — run these subagents IN THIS ORDER
Dispatch each as a subagent (Agent tool) with the company name and the working
directory path. Each stage depends on the previous one — run them
sequentially, not in parallel. Stop and report to the user if a stage fails.
- Generalist → writes
00_key_variables.md (key financial variables to
focus on for this company).
- Data Retriever → runs
tools/edgar_fetch.py to fetch CIK + latest 10-K +
companyfacts.json into raw/, and writes raw/meta.json. If the company
can't be resolved, stop and ask the user to disambiguate (use the candidate
list the tool prints).
- Data Scientist → writes
02_financials.json (cleaned metrics, ratios,
growth) from raw/companyfacts.json.
- Financial Analyst → writes
03_analysis.md (rigorous analysis grounded in
the cleaned data + 10-K).
- Report Writer → writes
report.html (the final ~2-page self-contained
website).
Use the matching subagent type for each stage: generalist, data-retriever,
data-scientist, financial-analyst, report-writer.
Output
The deliverable is output/<TICKER>/report.html. Report its path to the user and
offer to open it in a browser (e.g. open output/<TICKER>/report.html on macOS).
Notes
- All data is grounded in the company's latest 10-K and XBRL facts from SEC
EDGAR. There is no live market-price feed; valuation is framed qualitatively or
via per-share metrics.
- The EDGAR tool uses only the Python standard library — no
pip install needed.
SEC requires a descriptive User-Agent; the default is set in the tool and can
be overridden via the EDGAR_UA environment variable.
- Never fabricate financial figures. If data is missing, the report says so.
