The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts build_cards.py and make_graphic.py utilize subprocess.run to execute headless Chrome or Chromium. This is used to render HTML templates and Markdown into PNG images for report cards. The command is constructed using a list of predefined arguments and local file paths, which follows security best practices for subprocess execution.
[EXTERNAL_DOWNLOADS]: The skill fetches scientific paper PDFs and metadata from well-known and trusted external services, specifically arXiv, Crossref, Semantic Scholar, and OpenAlex. These network operations are core to the skill's functionality and target reputable academic infrastructure.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, untrusted data (academic papers) to generate summaries. This creates a theoretical surface for indirect prompt injection. However, the skill implements a rigorous 'fact ledger' (ledger.md) verification process and a detailed planning rubric (rubric.md) that instructs the agent to verify every claim against the original paper text, mitigating the risk of following hidden instructions in processed documents.

paper-report