paddleocr-doc-parsing
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes content from untrusted external documents (images, PDFs) provided by users. The extracted text is returned to the agent without specific delimiters or sanitization. While this creates a potential surface for indirect prompt injection, it is an inherent characteristic of document parsing skills and is managed by standard agent safety guardrails.\n- [COMMAND_EXECUTION]: The skill utilizes Python scripts to manage document parsing workflows, including image compression and PDF manipulation. These scripts interact with the local filesystem to read input documents and write results to the system temp directory.\n- [EXTERNAL_DOWNLOADS]: Document content is transmitted to a user-configured external API endpoint for processing. The skill ensures these communications are secure by using HTTPS and passing credentials via standard authorization headers.\n- [CREDENTIALS_UNSAFE]: The skill manages API access tokens through environment variables or a local .env file. It includes documentation advising users to configure these credentials securely within their host environment rather than sharing them in chat history.
Audit Metadata