The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill uses unvalidated user input from $ARGUMENTS (specifically strategy and symbol) to construct file system paths (backtesting/{strategy_name}/ and {symbol}_{strategy}_backtest.py) and to select template assets.
Ingestion points: The $ARGUMENTS variable parses user input which is directly interpolated into directory creation and file writing operations within SKILL.md.
Boundary markers: No delimiters or safety instructions are provided to ensure that inputs do not contain path traversal sequences (e.g., ../).
Capability inventory: The skill utilizes Write, Edit, and Bash tools, which allow for file system modifications and script execution.
Sanitization: There is no mention of sanitizing the $ARGUMENTS or validating that the resulting paths remain within the intended backtesting/ directory.
[DYNAMIC_EXECUTION]: The skill is designed to generate complete, executable Python scripts (.py files) based on user-provided parameters.
The generation process involves reading local assets and assembling code that includes external library calls and data fetching logic.
While this is the intended purpose of the backtesting skill, the generation of executable code from user-influenced strings increases the potential impact of injection attacks.
[COMMAND_EXECUTION]: The skill includes Bash in its allowed-tools and provides instructions that require directory creation and file manipulation.
This capability, combined with the dynamic generation of scripts, provides a significant surface for command execution if the agent is misled by malicious input provided in the $ARGUMENTS field.

backtest