Sec 10k Company Analysis

SEC 10-K Company Analysis

Use this skill to analyze one company from a SQLite SEC filings database and produce distinct, data-grounded QA pairs.

Inputs you need

• Company identifier: preferably CIK (or ticker/name if CIK is unavailable).
• Database location (or the active DB MCP connection).
• Target output count if provided; otherwise produce 8-12 distinct QA pairs.

Required workflow

1. Discover schema and confirm table/column names before deep queries.

• Always inspect available tables first.
• For financial_facts, confirm key columns such as:
  • fact_name, fact_value, unit, fiscal_year, fiscal_period, end_date, accession_number, form_type, filed_date, dimension_segment, dimension_geography.
• Never assume aliases like tag; use real column names.

2. Resolve the company identity.

• Query companies for base profile (name, SIC, description, fiscal year-end).
• If needed, query company_tickers for ticker mapping.

3. Build filing context.

• Query filings for recent records and isolate 10-K accession numbers.
• Keep a small set of recent annual filings (for trend analysis) plus the latest 10-K (for detail drills).

4. Discover available metrics before forcing templates.

• Enumerate candidate fact_name values for the company and latest 10-K.
• Revenue/profit labels vary by issuer; search for alternatives first, then lock in the best available tags.

5. Pull evidence in structured batches.

• Trend batch (multi-year): revenue, net income, assets, liabilities, equity, operating cash flow, financing/investing cash flow, debt, shares.
• Detail batch (latest filing): profitability, cost structure, liquidity, leverage, capital returns, leases, taxes, segment/geography if present.
• Use accession and date filters to avoid mixing inconsistent periods.

6. Draft QA pairs only from observed evidence.

• Each QA should cover a distinct analytical angle.
• Include concrete values/trends in the answer.
• Avoid speculation and avoid repeating the same thesis with different wording.

Query strategy patterns

Use patterns like these and adapt to actual schema:

• Company lookup

  • SELECT * FROM companies WHERE cik = '<CIK>'

• Recent annual filings

  • SELECT cik, form, filing_date, report_date, accession_number FROM filings WHERE cik = '<CIK>' AND form = '10-K' ORDER BY filing_date DESC LIMIT 10

• Metric discovery

  • SELECT DISTINCT fact_name FROM financial_facts WHERE cik = '<CIK>' AND form_type = '10-K' ORDER BY fact_name LIMIT 200

• Revenue alias discovery

  • SELECT DISTINCT fact_name FROM financial_facts WHERE cik = '<CIK>' AND form_type = '10-K' AND (fact_name LIKE '%Revenue%' OR fact_name LIKE '%Sales%' OR fact_name LIKE '%ContractWithCustomer%') ORDER BY fact_name

• Multi-year core metrics

  • SELECT fact_name, fact_value, unit, end_date, accession_number FROM financial_facts WHERE cik = '<CIK>' AND form_type = '10-K' AND fact_name IN (...) ORDER BY accession_number, fact_name, end_date

• Optional segment/geography coverage

  • SELECT DISTINCT dimension_segment, fact_name, fact_value, unit, end_date FROM financial_facts WHERE cik = '<CIK>' AND form_type = '10-K' AND dimension_segment IS NOT NULL AND dimension_segment != '' LIMIT 100
  • SELECT DISTINCT dimension_geography, fact_name, fact_value, unit, end_date FROM financial_facts WHERE cik = '<CIK>' AND form_type = '10-K' AND dimension_geography IS NOT NULL AND dimension_geography != '' LIMIT 100

Distinct QA angle checklist

Pick non-overlapping angles based on available evidence:

• Profitability trajectory and margin resilience
• Revenue growth/volatility drivers
• Cash flow quality vs accounting earnings
• Capital allocation (dividends, repurchases, capex)
• Balance sheet strength and debt servicing capacity
• Liquidity and working capital dynamics
• Cost structure shifts
• Share count / per-share implications
• Segment or geographic concentration (if present)
• Industry-specific risks visible in filings (e.g., lease burden, commodity cyclicality, impairment patterns)

Edge-case handling

• Missing expected metrics:
  • Discover alternates by fact_name search; do not invent absent fields.
• Empty result sets:
  • Relax one filter at a time (date, accession, strict metric list), then re-check.
• Mixed quarterly and annual facts:
  • Keep 10-K trend analysis annual-focused; avoid blending 10-Q values unless explicitly requested.
• Duplicated facts for same period:
  • Prefer consistent unit/date/accession combinations and document only stable comparisons.
• Tool/query errors:
  • Correct query using schema feedback and continue; do not repeat the same failing query.

Output format

For each QA pair:

• q: one analytical question with clear scope.
• a: concise answer grounded in retrieved facts (values, direction, period).

Quality bar:

• Evidence-grounded, specific, non-redundant.
• No unsupported claims.
• Answers should be interpretable without extra context.