The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill generates Python scripts dynamically, writing them to a temporary directory (tmp/magnific_upload.py) and executing them via the Bash tool to handle multi-step file uploads and downloads.
[INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface by ingesting local files and downloading content from external URLs returned by tool outputs. While intended for image data, this surface could be exploited if malicious instructions are embedded in the processed data.
Ingestion points: Local file content is read using pathlib.Path.read_bytes, and external results are fetched via urllib.request.urlretrieve using URLs from tool outputs.
Boundary markers: No explicit delimiters or instructions to ignore embedded content are used during data processing.
Capability inventory: The skill possesses Bash and Write capabilities for script execution, and uses the urllib library for network operations.
Sanitization: There is no evidence of sanitization or validation of the files or URLs processed by the generated scripts.
[COMMAND_EXECUTION]: The skill instructs the agent to use the Bash tool to execute system commands, including running Python scripts and performing image format conversions using the Pillow library.

magnific-local-upload