U.S. SEC Analysis

Use this skill to turn SEC EDGAR filings and XBRL facts into a reusable evidence pack before broader U.S. stock interpretation.

Language

• Default to Korean for user-facing output, headings, notes, and summaries unless the user explicitly asks for English.
• Keep official company names, ticker symbols, SEC form names, accounting labels, accession numbers, and quoted disclosure text exactly as disclosed when precision matters.

Quick Start

• Lock the exact company, ticker or CIK, target forms, and whether the user wants a filing-only pack or an input to a broader memo.
• Prefer the latest 10-K and 10-Q plus latest 8-K metadata for the default filing set.
• Use SEC primary sources first: data.sec.gov submissions, SEC XBRL companyfacts, and the filing's primary document in /Archives/edgar/data/.
• Declare a real User-Agent for SEC requests. Use SEC_USER_AGENT or pass --user-agent; keep request cadence below SEC fair-access limits.
• If the workspace is writable and the user wants a reusable artifact, write the result to analysis-example/us/<company>/sec-analysis.md.
• For long filings or memo-critical work, also create analysis-example/us/<company>/sec-reference.md and analysis-example/us/<company>/sec-cache.json so downstream work can cite the digest/cache without re-fetching the entire filing.
• If existing sec-analysis.md, sec-reference.md, or sec-cache.json files already exist for the company, read them before re-collecting and only refresh when the requested filing/date scope requires it.

Workflow

1. Lock scope. Confirm the ticker or CIK, target form set, latest versus specific period, and the key filing questions.
2. Collect SEC data. Run scripts/fetch-sec-edgar.js to resolve ticker-to-CIK, collect submissions, collect companyfacts, select latest forms, download primary filing HTML, and emit text plus JSON artifacts.
3. Parse filing sections. Run scripts/extract-sec-sections.js on the relevant latest-10k.txt or latest-10q.txt when the user needs MD&A, Risk Factors, controls, legal proceedings, or financial-statement sections.
4. Build a reusable reference. Run scripts/build-sec-reference.js to generate sec-reference.md and sec-cache.json from filing export, sections, and facts.
5. Write the analysis. Use the default SEC analysis shape from references/output-format.md, keep every material number source-mapped, and label gaps as not separately disclosed or needs_review.
6. Hand off cleanly. Use us-stock-analysis after this when the user wants thesis, valuation, catalysts, risks, and conclusion.

Read references/workflow.md for the SEC collection and extraction checklist. Read references/output-format.md for the default Korean evidence-pack shape. Read references/script-inputs.md before running the bundled scripts.

Bundled Scripts

• scripts/fetch-sec-edgar.js resolves ticker or CIK, fetches SEC submissions/companyfacts, downloads selected filing HTML, writes filing text files, and emits sec-filing-export.json plus sec-companyfacts.json.
• scripts/extract-sec-sections.js parses 10-K/10-Q Item headings and labels section coverage as parsed, partial, missing, or needs_review.
• scripts/build-sec-reference.js renders sec-reference.md and sec-cache.json from the fetch and section artifacts.
• Run all bundled scripts with node.

Operating Rules

• Do not invent revenue mix, customer concentration, segment profitability, governance facts, or valuation metrics that are not in the current SEC source set.
• Do not derive standalone quarters from year-to-date XBRL facts unless the user explicitly asks and you have the matching prior cumulative fact; v1 should preserve XBRL start/end period facts as disclosed.
• Preserve accession number, form, filed date, period start/end, unit, taxonomy, and source URL for each material XBRL fact.
• Treat SEC XBRL companyfacts as standardized fact evidence, not as a complete substitute for reading the filing text.
• If a company uses only custom taxonomy for a memo-critical metric and companyfacts lacks the standard concept, mark it as not separately disclosed in standard SEC companyfacts and inspect the filing text.
• Keep latest 8-K work metadata-first unless the user asks to parse 8-K items.
• If a filing section parse is missing or needs_review, do not convert that into not separately disclosed.
• Keep SEC fair access explicit: declared User-Agent, efficient downloads, cache reuse, and request cadence below 10 requests/second.

Source Priority

1. SEC filing primary document and inline XBRL filing text
2. SEC data.sec.gov/submissions/CIK##########.json
3. SEC data.sec.gov/api/xbrl/companyfacts/CIK##########.json
4. Company IR earnings releases or decks for management commentary
5. Reputable news/transcripts only for context, not as a substitute for filing facts

Minimum Output Standard

• Scope with ticker, CIK, company name, forms, accession numbers, filing dates, and period dates
• Latest 10-K and 10-Q filing links, plus latest 8-K metadata when available
• XBRL fact table for revenue, gross profit, operating income, net income, diluted EPS, cash, assets, liabilities, equity, operating cash flow, and capex when disclosed in standard concepts
• MD&A, Risk Factors, financial statements, controls, and other requested filing sections with coverage status
• Clear not separately disclosed and needs_review notes
• Source Map table another analyst can audit quickly
• Reusable sec-reference.md and sec-cache.json for long or memo-critical filing work

Handoff Guidance

• Use us-stock-analysis after this when the user wants a final investment memo, valuation, catalysts, risks, or comparison.
• If the user's question is mainly "what does the SEC filing actually say", keep the output in SEC analysis mode and stop short of a thesis.
• If an existing sec-reference.md or sec-cache.json covers the same accession numbers, cite and update that artifact instead of re-downloading unchanged filings.