The Agent Skills Directory

[PROMPT_INJECTION]: The skill directly uses command-line arguments ($0, $1, etc.) to control file system operations and code generation without sanitization. Evidence: $ARGUMENTS are parsed and used directly to construct directory and file paths.
[DATA_EXFILTRATION]: Path traversal vulnerability exists in the template selection logic, where a manipulated template name could be used to read files outside the intended assets directory. Evidence: The instruction to read from 'algo-expert/rules/assets//strategy.py' allows an attacker to provide a value like '../../../../etc/passwd'.
[COMMAND_EXECUTION]: The skill instructs the agent to create directories and execute Python scripts at paths constructed from user input, which can be exploited if path traversal occurs. Evidence: Instructions to run 'python strategies//strategy.py' where is derived from unsanitized arguments.

algo-strategy