AGENT SKILL

13f-analyzer

Perform various data analysis on SEC 13-F and obtain some insights of fund activities such as number of holdings, AUM, and change of holdings between two quarters.

License 许可
Apache-2.0
最近更新
2026-11-14
Stars
★ 0
Forks
1

01 前言 Front Matter

name
13f-analyzer
description
Perform various data analysis on SEC 13-F and obtain some insights of fund activities such as number of holdings, AUM, and change of holdings between two quarters.

02 资源文件 2

03 SKILL.md 正文

Overview

The helper programs are installed under /skills/13f-analyzer/scripts. Run them from that absolute path; the task workspace is normally /root, so scripts/... and /root/scripts/... are not equivalent paths. If a helper cannot be found, check /skills/13f-analyzer/scripts before writing a replacement analysis pipeline.

The helpers define the stock-selection and aggregation semantics for this skill. Do not silently broaden their stock universe when reproducing their results.

Analyze the holding summary of a particular fund in one quarter:

python3 /skills/13f-analyzer/scripts/one_fund_analysis.py \
    --accession_number ... \
    --quarter 2025-q2

Use this helper to read the fund’s quarter data and report totals such as AUM (sum of VALUE for that accession number in that quarter).

For questions asking “how many stocks” the fund holds in a quarter, use the count of distinct stock positions (unique CUSIPs) after the helper’s stock filter and CUSIP grouping (i.e., count rows of the grouped stock table), e.g.:

python3 - <<'PY'
import sys
sys.path.append("/skills/13f-analyzer/scripts")
import one_fund_analysis as ofa

quarter = "2025-q2"
accession = "ACCESSION_NUMBER_HERE"

stock = ofa.read_one_quarter_data(accession, quarter)  # grouped by CUSIP
print(stock.shape[0])
PY

Analyze the change of holdings of a particular fund in between two quarters:

python3 /skills/13f-analyzer/scripts/one_fund_analysis.py \
    --quarter 2025-q3 \
    --accession_number <accession number assigned in q3> \
    --baseline_quarter 2025-q2 \
    --baseline_accession_number <accession number assigned in q2>

This script will print out the dynamic changes of holdings from 2025-q2 to 2025-q3. Such as newly purchased stocks ranked by notional value, and newly sold stocks ranked by notional value.

For a request ranked by dollar-value increase, take the requested number of CUSIPs from the beginning of the helper's Top Buys output, preserving order. Do not rank by share-count change, percentage change, or an unfiltered INFOTABLE.

If the comparison must be reproduced in custom code, preserve the helper's pipeline exactly:

  1. In each quarter, select the requested accession number and apply the same stock TITLEOFCLASS filter used by the helper.
  2. Group each filtered quarter by CUSIP and sum VALUE.
  3. Outer-join the two quarters and fill missing current/baseline values with 0.
  4. Compute ABS_CHANGE = current VALUE - baseline VALUE.
  5. Keep positive changes and sort ABS_CHANGE descending.

Before returning an answer, verify that the requested top-k contains exactly k unique CUSIP strings in the printed order.

Analyze which funds hold a stock to the most extent

Compute top holders from the quarter-scoped data, then map accession numbers to fund manager names via the same quarter’s COVERPAGE.tsv:

python3 - <<'PY'
import pandas as pd

quarter = "2025-q3"
cusip = "CUSIP_HERE".upper()
topk = 10

it = pd.read_csv(f"/root/{quarter}/INFOTABLE.tsv", sep="\t", dtype=str)
it["CUSIP"] = it["CUSIP"].str.upper()
it["VALUE"] = it["VALUE"].astype(float)

top = (
    it[it["CUSIP"] == cusip]
      .groupby("ACCESSION_NUMBER")["VALUE"].sum()
      .sort_values(ascending=False)
      .head(topk)
      .reset_index(name="TOTAL_VALUE")
)

cp = pd.read_csv(f"/root/{quarter}/COVERPAGE.tsv", sep="\t", dtype=str)[
    ["ACCESSION_NUMBER", "FILINGMANAGER_NAME", "ISAMENDMENT"]
]
cp = cp[cp["ISAMENDMENT"] == "N"].drop_duplicates("ACCESSION_NUMBER")

out = top.merge(cp, on="ACCESSION_NUMBER", how="left")
print(out.to_string(index=False))
PY

This prints the top holders (by summed VALUE) and their FILINGMANAGER_NAME.

04 同仓库更多 Skills