The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze and modify project-specific metadata files such as MANIFEST.MF, plugin.xml, and build.properties. While this introduces an indirect prompt injection surface where malicious instructions could be embedded in the data being processed, the skill provides a standard Python-based checker (scripts/check_bundle.py) that uses safe parsing methods (regex and xml.etree.ElementTree). The instructions also emphasize treating workspace files as data and avoiding testing against the user's primary workspace. Ingestion points: MANIFEST.MF, plugin.xml, build.properties, and Java source files are read from the local project directory. Boundary markers: The instructions include clear warnings to treat workspace files as data and to avoid installing into the user's own IDE. Capability inventory: The skill can execute shell commands via verify.sh (running Maven, Java, Javac) and write to the local filesystem (markers, preferences, and project files). Sanitization: Uses standard Java and Python libraries for parsing manifest and XML files.\n- [COMMAND_EXECUTION]: The provided assets/examples/verify.sh script executes various build and test commands, including mvn, java, javac, and python3. These are standard development operations for the skill's stated purpose of building and testing Eclipse plugins.\n- [EXTERNAL_DOWNLOADS]: The skill's documentation and verification scripts reference external resources, primarily from download.eclipse.org and docs.osgi.org. These are well-known, official sources for the Eclipse ecosystem. The automated phishing alert for the OSGi documentation link is evaluated as a false positive given its role as the official standards body for the underlying technology.

develop-eclipse-ide-plugins