The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill interpolates user-provided ticker symbols directly into an LLM prompt to fetch market prices. An attacker could provide a ticker name containing instructions to manipulate the LLM's JSON output.
Ingestion points: Ticker symbols provided via the add or update commands in scripts/portfolio.py are used in the fetch_prices function.
Boundary markers: The prompt in fetch_prices lacks clear delimiters (like XML tags or triple backticks) around the variable list of tickers, making it easier for an injected string to blend into the instructions.
Capability inventory: The skill uses the openai library to make network requests to api.aisa.one and writes portfolio data to a local JSON file.
Sanitization: Ticker symbols are converted to uppercase using .upper(), but no validation is performed to ensure the input is a valid alphanumeric ticker symbol, allowing for the inclusion of control characters or command-like text.

stock-portfolio