The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from an external CSV file, creating a surface for indirect prompt injection where malicious instructions in the file could influence agent behavior.
Ingestion points: The skill reads ticker data, action flags, and free-text triggers from stocks/portfolio-review.csv (located four levels above the script directory).
Boundary markers: Absent. There are no delimiters or instructions provided to the agent to treat the contents of the CSV as untrusted or to ignore embedded instructions.
Capability inventory: The skill executes a local Python script, performs network requests to fetch market data, and reads/writes files to the local file system.
Sanitization: Absent. While the Python script uses regex to identify price levels, it extracts the surrounding text and includes it verbatim in a markdown report, which the agent is then instructed to read and summarize.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill accesses user data files outside of its immediate skill directory. The script defaults to a path four levels above its location (../../../../stocks/portfolio-review.csv) to read portfolio details. It also transmits ticker symbols to Yahoo Finance to retrieve current pricing information.
[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill relies on the yfinance library to perform network requests and retrieve live market data from Yahoo Finance, a well-known financial data provider.

portfolio-monitor