doubao-earnings-analysis

Analyze listed-company earnings reports with causal attribution and source-bound fact verification.

2|Updated Aug 9, 2026
One-click install
npx skills add https://github.com/DeepJH/doubao-skill-and-info --skill doubao-earnings-analysis-deepjh
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: doubao-earnings-analysis
Source: https://github.com/DeepJH/doubao-skill-and-info/tree/main/skills/doubao-earnings-analysis
Command: npx skills add https://github.com/DeepJH/doubao-skill-and-info --skill doubao-earnings-analysis-deepjh

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires python-docx, and includes scripts (resource) and references (resource) components.

What problem does it solve? It turns raw earnings report numbers into causal explanations—why revenue, margins, or cash flow changed and whether the company can reverse the change—while enforcing that every figure in the output is traceable to a named, dated source. ## Core Features & Use Cases - Two analysis modes: A report mode for full earnings reviews and a focused mode for specific questions like "why did gross margin fall", routed automatically by question scope. - Fact verification pipeline: A facts.json registry plus Python scripts (check_facts.py, lint_report.py, finalize_report.py) that validate claims, block unsupported "beat/miss expectations" wording, and convert internal fact bindings into numbered source citations. - Market-aware conventions: Separate disclosure calibers for A-shares, Hong Kong stocks, and US-listed Chinese companies, with industry playbooks covering product, financial, project-delivery, and recurring-revenue business models. - Use Case: Ask "why did Company X's Q3 gross margin decline" and receive a sourced causal analysis with confidence-graded conclusions, a follow-up watchlist, and a Feishu document deliverable. ## Quick Start Ask the assistant to analyze why a specific listed company's gross margin declined in its latest quarterly report and deliver the result as a Feishu document.

Frequently Asked Questions about doubao-earnings-analysis

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I analyze why a company's gross margin declined in an earnings report?

Ask a focused question naming the company, report period, and metric, such as "why did gross margin fall in Q3". The skill routes scoped questions to focused mode, decomposes the change through cost and product-mix hypotheses, and pushes each explanation to competitive-constraint depth.

What stock markets does this earnings analysis skill support?

It covers A-shares, Hong Kong stocks, and US-listed Chinese companies. Each market uses its own profit terminology: A-shares use net profit attributable to parent and deducted figures, Hong Kong uses shareholder-attributable and adjusted profit, and US listings use GAAP versus non-GAAP.

How does the skill verify financial numbers before publishing?

Every key figure is registered in a facts.json file with source, date, and usage type, then validated by check_facts.py and lint_report.py. Claims from brokers or media cannot be written as company-disclosed facts, and beat-or-miss wording requires a documented consensus anchor.

Can the skill give stock ratings, target prices, or valuation judgments?

No. The skill explicitly prohibits outputting ratings, target prices, and valuation multiples. It focuses on causal interpretation of reported results and provides a follow-up watchlist of verifiable public indicators instead.

What is the difference between report mode and focused mode?

Report mode handles broad requests covering an entire earnings release, such as "how was this quarter". Focused mode handles questions scoped to one or two variables, like "why did R&D spending jump". The routing is decided by the scope of the user's question.

What happens when consensus estimates or data are unavailable?

The skill degrades gracefully: it declares that consensus was not obtained and anchors against company guidance or historical trends instead. Unverifiable single explanations are labeled as possibilities with explicit follow-up indicators rather than stated as conclusions.