The Agent Skills Directory

[SAFE]: The skill's primary function is to process local PDF files and generate summaries. The implementation utilizes the official google-genai SDK and standard PDF processing libraries like PyMuPDF.
[CREDENTIALS_UNSAFE]: Security best practices are followed for credential management; the GEMINI_API_KEY is accessed via environment variables, and no hardcoded secrets were found in the scripts or metadata.
[DATA_EXFILTRATION]: Network operations are limited to communicating with Google's Gemini API endpoints for document analysis. This behavior is consistent with the skill's stated purpose, and no data is sent to untrusted third-party domains.
[REMOTE_CODE_EXECUTION]: The skill does not contain any remote code execution patterns, dangerous shell command invocations, or calls to eval() or exec(). Logic is contained within local Python scripts and standard templates.
[PROMPT_INJECTION]: The skill's instructions and templates are designed for structured summarization. There are no attempts to bypass safety filters, jailbreak the model, or override system instructions.
[INDIRECT_PROMPT_INJECTION]: While the skill processes untrusted PDF data (which is a common surface for indirect injection), it employs structured templates and multi-modal input handling (sending PDF bytes directly to Gemini), which significantly mitigates the risk of the model inadvertently executing instructions found within the document text.

gemini-pdf-summary