The Agent Skills Directory

[COMMAND_EXECUTION]: The skill scripts utilize subprocess.run to interface with system binaries such as convert, tesseract, and exiftool. These calls use argument lists rather than raw strings, which effectively prevents shell injection by not invoking a shell to interpret the command. This is a secure implementation of system command wrappers.
[PROMPT_INJECTION]: The skill processes untrusted input through OCR and metadata extraction, creating an ingestion surface for indirect prompt injection.
Ingestion points: scripts/ocr.py and scripts/read_exif.py ingest data from external image files, which could contain malicious text or metadata.
Boundary markers: The scripts return raw text or JSON data to the agent without explicit delimiters or instructions to ignore embedded commands.
Capability inventory: The skill is limited to reading/writing image files and executing specific image processing binaries via subprocess.run.
Sanitization: Shell injection is mitigated by the use of list-based arguments in all subprocess calls. The skill does not perform network operations or access sensitive credentials, limiting the impact of any potential injection.

image