Trader Analyst skill

trader-analyst

Earnings season moves stocks, and the press release is written to put the best number first. This skill reads the numbers the company actually filed with the SEC, lines them up quarter by quarter, and says plainly what changed.

Tool in this folder (standard library Python, no keys):

python3 "${CLAUDE_SKILL_DIR}/facts.py" --help

It needs the SEC User-Agent (sec_user_agent in trader/config.json, or SEC_USER_AGENT): "Your Name you@example.com" with the user's real name and email. It refuses to call SEC without one.

What it reads and writes

• Reads https://data.sec.gov/api/xbrl/companyfacts/CIK##########.json (every number the company has tagged in its filings), the company's submissions list (for the newest 10-Q or 10-K and the newest 8-K with item 2.02, "Results of Operations") and that 8-K's folder to find the press release (the EX-99.1 exhibit).
• Writes trader/metrics.json with --metrics: one row per company (TTM revenue, net income and diluted EPS, the latest margins and growth) for /trader-screener. metrics.example.json in this folder is a real run for AAPL, AXP, GOOGL, KO and BAC on 2026-10-13.

The rules

• Filed numbers first. The table comes from XBRL facts the company filed. Use the press release for context (guidance, segment colour), and say which numbers came from where.
• Compare like with like. Growth is against the same quarter a year earlier. Quarter-on-quarter changes are shown but many businesses are seasonal (Apple's December quarter is always its biggest).
• Say what is derived. Companies do not file a Q4 10-Q. The fourth quarter is the 10-K year minus the three quarters inside it and is marked "derived"; derived EPS is approximate.
• Say when the data lags. SEC's facts API can trail a filing. When the newest 10-Q is later than the newest quarter in the facts, the tool prints a NOTE with the link: read the 10-Q itself before summarising.
• Flag one-offs. When net income is above operating income, gains outside the business (investment marks, tax items) are inside it. The tool says so. Banks and insurers have no gross margin: blank, not zero.

Steps

1. Run it for the companies in question (or every ticker in trader/targets.json):

   python3 "${CLAUDE_SKILL_DIR}/facts.py" AAPL
   python3 "${CLAUDE_SKILL_DIR}/facts.py" AAPL AXP GOOGL KO BAC --metrics trader/metrics.json
   

   --quarters 8 shows two years, --json everything machine-readable, --cik a company without a ticker.

2. Open the latest earnings press release (the EX-99 link) and the 10-Q when the numbers need context, and read them.

3. Write the summary: one line on the quarter (revenue, EPS, growth), one on margins, one on anything unusual (a one-off, a derived quarter, data that lags), and the links. Plain words, real numbers, no adjectives the numbers do not support.

Example: the reel's top holding

facts.py AAPL (2026-10-13): quarter ending 2026-06-27 (fiscal Q3 2026), revenue $109.42B, +16.4% against a year earlier, net income $29.79B (+27.1%), diluted EPS $2.02 (+28.7%); gross margin 50.1% (+0.8 points on the quarter before); last four quarters $466.8B revenue and $8.71 EPS. Press release: the EX-99.1 of the 8-K filed 2026-07-30. For Coca-Cola the tool notes that SEC's facts stop at 2026-04-03 while the 10-Q for 2026-07-03 is already filed, so the summary reads that 10-Q directly.

What it never does

• Never says a stock is cheap, expensive, a buy or a sell. It reports what the company filed.
• Never invents a number the filings do not have, or estimates a quarter it cannot derive.
• Never calls SEC without the User-Agent, and stays well under 10 requests a second.

Safety

Education and tooling, not financial advice. This skill only reads filings. Nothing places an order except /trader-broker, after /trader-risk passes and the user types an explicit yes for that order.