dwg-extraction

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly ingests and parses untrusted third-party files: SKILL.md and scripts/parse_dxf.py require reading user-/third-party .dwg→.dxf content (parsing TEXT/MTEXT/ATTRIB values) and scripts/compile_libredwg.sh clones code from the public GitHub repo (https://github.com/LibreDWG/libredwg.git), so arbitrary external text embedded in those DWG/DXF files or fetched source could influence extraction results and downstream actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly runs git clone during runtime to fetch and compile remote source from https://github.com/LibreDWG/libredwg.git, which is executed (producing/ running the dwg2dxf binary) and is a required dependency for the pipeline.