The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Python scripts including scripts/pdf_page_count.py, scripts/pdf_to_images.py, and scripts/render_pdf.py to execute local system tools. Evidence: The scripts invoke external binaries such as pdfinfo, pdftoppm, magick, and google-chrome via subprocess.run to process PDF files and render HTML. These calls use list-based arguments which mitigate common shell injection vulnerabilities.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which introduces an attack surface for indirect instructions. 1. Ingestion points: Job descriptions (JD) provided as text and content from user-uploaded resume PDFs (processed via image understanding) as described in SKILL.md. 2. Boundary markers: The skill instructions do not specify any delimiters or ignore-instructions warnings for isolating external content. 3. Capability inventory: The agent possesses capabilities for local file system writes and execution of PDF processing scripts. 4. Sanitization: No explicit sanitization or content filtering is implemented for the external job descriptions or resume data before processing.
[EXTERNAL_DOWNLOADS]: Resume templates reference external assets for layout and placeholders. Evidence: Templates in assets/template_refs/html/ load images from Alibaba Cloud OSS (oss-cn-shanghai.aliyuncs.com) and utilize well-known CDN services like cdnjs and Google Fonts for CSS and typography. These references are consistent with the intended functionality and originate from well-known infrastructure.

resume-master