docling-document-intelligence

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly shows passing a PDF password as a command-line argument (--pdf-password PASSWORD), which requires embedding a secret verbatim in generated commands (and the Python VLM API section allows endpoint/credential configuration), creating an exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.90). The skill's required workflow explicitly accepts and fetches documents by URL (see SKILL.md Step 1 and CLI examples like docling https://example.com/a.pdf, and EXAMPLE.md usage such as converting arXiv URLs), then parses and analyzes that third‑party content (convert, chunk for RAG, and drive refinement/evaluation steps), meaning untrusted public documents can be ingested and materially influence agent decisions and tool use.