The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external websites (Shiller data via Yale, FRED CSVs, and Yahoo Finance) which could theoretically be manipulated to include malicious instructions processed by the AI.
Ingestion points: scripts/fetch_valuation_data.py (via requests and pd.read_excel/pd.read_csv) and scripts/valuation_percentile.py.
Boundary markers: The skill does not implement delimiters or 'ignore embedded instructions' warnings when processing this external data.
Capability inventory: The skill has file-write capabilities (saving CSV, JSON, and PNG reports) and executes subprocesses for visualization.
Sanitization: There is no evidence of sanitization or filtering of the content retrieved from the internet before it is potentially used in the agent's context.
[EXTERNAL_DOWNLOADS]: The skill fetches financial data from external sources at runtime.
Fetches the Shiller CAPE dataset from Yale University (http://www.econ.yale.edu/~shiller/data/ie_data.xls). Note: This source uses an unencrypted HTTP connection.
Fetches economic data from the Federal Reserve Bank of St. Louis (FRED) via its CSV API (https://fred.stlouisfed.org/graph/fredgraph.csv).
Retrieves stock data through the yfinance library, which interacts with Yahoo Finance APIs.
[COMMAND_EXECUTION]: The skill instructions and workflows require the execution of Python scripts (valuation_percentile.py, visualize_valuation.py) to perform calculations and generate charts. This is the intended functionality of the skill.

detect-us-equity-valuation-percentile-extreme