building-java-knowledge-graph

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). Outsider free text can enter the LLM context via the project’s own source files and config files (e.g., *.java/*.kt/*.scala/*.groovy, application*.properties/yaml) that the script reads at runtime and then serializes into knowledge-graph.json (nodes/edges/annotations/properties), which downstream agents may feed into the LLM.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). Yes — scripts/install_grammars.py performs git clone and builds tree-sitter grammars at setup/runtime, fetching and executing remote code from the listed GitHub repositories (https://github.com/tree-sitter/tree-sitter-java, https://github.com/fwcd/tree-sitter-kotlin, https://github.com/tree-sitter/tree-sitter-scala, https://github.com/murtaza64/tree-sitter-groovy), and the resulting artifacts (languages.so) are required by scripts/build_knowledge_graph.py.