The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/exr.py executes system commands via subprocess.call() to wrap OpenEXR utilities like exrheader, exrinfo, and exrmaketiled. Evidence:
- File: scripts/exr.py contains multiple calls to subprocess.call(cmd) where cmd is a list of arguments.
- The use of argument lists instead of raw shell strings follows best practices to prevent shell injection, although it still facilitates the execution of local system commands based on user input.
[PROMPT_INJECTION]: The skill processes metadata from external OpenEXR files, which can lead to indirect prompt injection if a malicious file contains instructions designed to override agent behavior. Evidence:
- Ingestion points: Functions in scripts/exr.py such as cmd_header, cmd_info, and cmd_stdattr read and display header attributes from .exr files provided to the agent.
- Boundary markers: The skill does not use specific delimiters or instructions (e.g., 'ignore embedded instructions') when outputting file metadata to the agent context.
- Capability inventory: The skill allows the agent to execute shell commands (via the wrapper) and manipulate files on the local system.
- Sanitization: No sanitization, validation, or escaping is performed on the metadata values extracted from the OpenEXR files before they are presented to the agent.

vfx-openexr