The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses string interpolation to build shell commands (find) and Python scripts (python3 -c "..."). It instructs the agent to replace placeholders like INPUT_PATH, OUTPUT_PATH, and INPUT_FILENAME with user-supplied values. This pattern is vulnerable to command injection if the input values are not properly sanitized or escaped before being placed into the script templates.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data including user input from $@, file names found on the local filesystem, and state information from .geoai-skills/state.json without sufficient boundary markers or sanitization logic.
Ingestion points: User arguments ($@), file paths from find output, and paths/metadata from state.json (SKILL.md).
Boundary markers: None present. There are no instructions for the agent to treat interpolated data as untrusted or to use specific delimiters.
Capability inventory: Shell execution (Bash) and Python script execution (python3 -c).
Sanitization: None present. The skill relies on simple string replacement for parameters like MINX and INPUT_PATH.

process-raster