The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill initializes a connection to the ImageJ2/Fiji ecosystem by fetching resources from official repositories.
Evidence: The imagej.init("sc.fiji:fiji") command automatically downloads the required Fiji distribution (approximately 500 MB) from established software mirrors on the first run if a local installation is not detected.
[DYNAMIC_EXECUTION]: The skill utilizes dynamic code evaluation and class loading to interface between Python and the Java Virtual Machine (JVM).
Evidence: The skill uses ij.macro.run() and ij.macro.runMacroFile() to execute ImageJ Macro Language (IJM) code strings and external script files within the embedded JVM.
Evidence: It employs jnius.autoclass() to dynamically reference and load Java classes at runtime, which is a standard requirement for PyImageJ functionality.
[INDIRECT_PROMPT_INJECTION]: The skill processes external macro files and image data, which introduces an attack surface for untrusted instructions if files are sourced from malicious origins.
Ingestion points: Data and logic enter the agent context through ij.io().open(path) for image hyperstacks and ij.macro.runMacroFile(path) for processing scripts in SKILL.md.
Boundary markers: The provided examples do not use specific delimiters or protective instructions when executing external macro files.
Capability inventory: The skill provides extensive capabilities including file system access (ij.io().save, df.to_csv), macro execution, and JVM-level operations.
Sanitization: There is no evidence of sanitization for macro content or file paths in the documented examples.

pyimagej-fiji-bridge