Sec13f data format

Skill cxcscmu/SkillLearnBench/skills/b1-one-shot-claude-haiku-4-5/financial-analysis/sec13f-data-format

[COLM'26] SkillLearnBench is the first benchmark for evaluating continual learning methods that automatically generate agent skills.

Install

npx -y skills add cxcscmu/SkillLearnBench --skill sec13f-data-format

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

What its author says it does

Copied from the file, not written here

Understanding SEC 13-F filing data structure, TSV format, and key tables for hedge fund analysis

SKILL.md

4.7 KB, as published. Nobody here has run it

SEC 13-F Data Format and Structure

Overview

SEC 13-F filings provide quarterly snapshots of large institutional investment holdings. The data is distributed as tab-separated value (TSV) files organized by reporting quarters.

Key Data Files

COVERPAGE.tsv

Purpose: Fund/Manager information and filing metadata Key Columns:

ACCESSION_NUMBER: Unique identifier for the filing (use to link with holdings)
FILINGMANAGER_NAME: Name of the fund or investment manager
REPORTCALENDARORQUARTER: Reporting date (e.g., "30-JUN-2025")
DATEREPORTED: When the filing was reported
REPORTTYPE: Type of 13-F report

Usage Example:

import pandas as pd

# Load fund information
coverpage = pd.read_csv('/root/2025-q3/COVERPAGE.tsv', sep='\t')

# Find a specific fund by name
fund_info = coverpage[coverpage['FILINGMANAGER_NAME'].str.contains('Renaissance', case=False, na=False)]
accession_number = fund_info['ACCESSION_NUMBER'].values[0]

INFOTABLE.tsv

Purpose: Detailed holdings data for all funds Key Columns:

ACCESSION_NUMBER: Links to the fund (from COVERPAGE)
NAMEOFISSUER: Company name of the stock
CUSIP: Committee on Uniform Security Identification Procedures code (unique stock identifier)
VALUE: Market value of holdings in thousands (USD)
SSHPRNAMT: Number of shares held (integer)
SSHPRNAMTTYPE: Share amount type (usually "SH" for shares)

Usage Example:

# Load all holdings
holdings = pd.read_csv('/root/2025-q3/INFOTABLE.tsv', sep='\t')

# Get holdings for a specific fund
fund_holdings = holdings[holdings['ACCESSION_NUMBER'] == 'specific_accession_number']

# Get count of stocks held
stock_count = len(fund_holdings)

# Find a specific stock
palantir = holdings[holdings['NAMEOFISSUER'].str.contains('PALANTIR', case=False, na=False)]

SUMMARYPAGE.tsv

Purpose: Summary-level information per fund Key Columns:

ACCESSION_NUMBER: Fund identifier
TABLE_OF_CONTENTS: Summary metadata
Other aggregate information

Data Characteristics

Value Representation

All monetary values in INFOTABLE are in thousands of USD
To get actual dollar value: actual_value = VALUE_FROM_TSV * 1000
Share counts are exact integers

Date Formats

Dates typically in format "DD-MMM-YYYY" (e.g., "30-JUN-2025")
Convert using: pd.to_datetime(df['date_column'], format='%d-%b-%Y')

Data Types

# Common data type conversions
df['VALUE'] = pd.to_numeric(df['VALUE'], errors='coerce')
df['SSHPRNAMT'] = pd.to_numeric(df['SSHPRNAMT'], errors='coerce')
df['CUSIP'] = df['CUSIP'].astype(str)

Loading and Initial Data Exploration

import pandas as pd

def load_13f_data(quarter_path):
    """Load all 13-F data for a quarter"""
    return {
        'coverpage': pd.read_csv(f'{quarter_path}/COVERPAGE.tsv', sep='\t'),
        'infotable': pd.read_csv(f'{quarter_path}/INFOTABLE.tsv', sep='\t'),
        'summarypage': pd.read_csv(f'{quarter_path}/SUMMARYPAGE.tsv', sep='\t')
    }

# Usage
q2_data = load_13f_data('/root/2025-q2')
q3_data = load_13f_data('/root/2025-q3')

# Find number of unique funds
num_funds = q3_data['coverpage']['ACCESSION_NUMBER'].nunique()
print(f"Number of unique funds in Q3: {num_funds}")

Common Analysis Patterns

Get Fund AUM (Assets Under Management)

AUM is typically found in SUMMARYPAGE.tsv:

summarypage = q3_data['summarypage']
# Look for AUM-related fields (may vary by filing)

Get Fund Holdings Count

fund_accession = 'specific_accession_number'
fund_holdings = q3_data['infotable'][q3_data['infotable']['ACCESSION_NUMBER'] == fund_accession]
num_holdings = len(fund_holdings)

Cross-Quarter Comparison

# Get same fund's holdings from two quarters
q2_holdings = q2_data['infotable'][q2_data['infotable']['ACCESSION_NUMBER'] == accession]
q3_holdings = q3_data['infotable'][q3_data['infotable']['ACCESSION_NUMBER'] == accession]

# Merge and compare
comparison = pd.merge(
    q2_holdings[['CUSIP', 'VALUE', 'SSHPRNAMT', 'NAMEOFISSUER']],
    q3_holdings[['CUSIP', 'VALUE', 'SSHPRNAMT']],
    on='CUSIP',
    suffixes=('_q2', '_q3'),
    how='outer'
).fillna(0)

# Calculate changes
comparison['value_change'] = comparison['VALUE_q3'] - comparison['VALUE_q2']
comparison = comparison.sort_values('value_change', ascending=False)

Notes

ACCESSION_NUMBER is the critical linking key between all tables
CUSIP is the universal stock identifier
Multiple funds can have the same stock in their portfolios
Data is case-sensitive in some fields, use .str.lower() or str.contains(..., case=False) for case-insensitive matching