The Agent Skills Directory

[PROMPT_INJECTION] (HIGH): Indirect Prompt Injection surface (Category 8).
Ingestion points: scripts/code_formatter.py reads file content from user-specified paths using open(path, 'r').
Boundary markers: Absent. There are no delimiters or instructions to ignore embedded commands within the files being formatted.
Capability inventory: The skill possesses file write capabilities, using open(path, 'w') to overwrite files on the local filesystem.
Sanitization: Absent. The tool does not sanitize or filter the content for potential AI instructions hidden in code comments or strings.
[DATA_EXFILTRATION] (HIGH): Potential for Path Traversal.
The script scripts/code_formatter.py accepts arbitrary file paths via command-line arguments and performs read and write operations on them. It lacks validation logic to ensure paths are restricted to safe or intended directories, potentially allowing an attacker to manipulate the agent into reading or corrupting sensitive files (e.g., ../../.env).
[COMMAND_EXECUTION] (LOW): The skill requires executing a Python script in a shell environment, which is the intended design for this tool.

code-formatter