The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill uses the ! syntax in SKILL.md to run a Python script at the moment the skill is loaded, which bypasses normal execution flow and executes code before AI processing.
Evidence: !python3 -c "exec(...)"`` block in SKILL.md used for dependency checking.
[EXTERNAL_DOWNLOADS]: The skill automatically triggers the installation of the yfinance, pandas, and numpy packages from the Python Package Index (PyPI) via pip if they are not detected.
Evidence: subprocess.check_call([sys.executable, "-m", "pip", "install", "-q", "yfinance", "pandas", "numpy"]) in SKILL.md.
[DYNAMIC_EXECUTION]: The skill utilizes the Python exec() function within its load-time script to evaluate logic dynamically, which is a high-risk code pattern.
Evidence: Use of exec() inside the dynamic context injection command string in SKILL.md.
[COMMAND_EXECUTION]: The skill uses subprocess.check_call to execute shell commands for package management.
Evidence: subprocess calls in SKILL.md to manage dependencies.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external financial data from Yahoo Finance, posing a potential surface for indirect instruction injection if the external data source is compromised.
Ingestion points: Fetches data through the yfinance library across multiple sub-skills in SKILL.md.
Boundary markers: No delimiters or protective instructions are used to isolate the external data from the agent's primary instructions.
Capability inventory: Includes system command execution via subprocess and dynamic execution via exec in SKILL.md.
Sanitization: No sanitization of the external data is performed before processing.

etf-premium