financial-deep-research — Claude Code skill: what it does

What financial-deep-research does

The skill runs research through an eight-phase pipeline: scope the question, plan the strategy, retrieve data in parallel, triangulate every claim against three or more sources, synthesize, critique, refine, and package. You pick a mode — quick, standard, deep or ultradeep — and standard is the default for most queries.

Inside are scripts/validate_report.py for checking the finished report, templates/report_template.md, and reference/methodology.md, which the skill loads phase by phase instead of all at once. Every financial claim has to cite a specific source, and inferences get marked separately from facts pulled out of filings. Simple stock quotes and basic company lookups are out of scope; the skill tells you to stop and use plain search instead.

It runs autonomously and rarely asks clarifying questions — only when the request contradicts itself or makes no sense. Jurisdiction defaults to US unless you say otherwise, so check that if you need another market.