The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches research paper metadata and abstracts from the official APIs of well-known academic services, including arXiv, Semantic Scholar, OpenAlex, CrossRef, and PubMed. These are trusted repositories for scientific data.
[COMMAND_EXECUTION]: The skill uses the generate_pdf.py script to invoke system tools like pandoc and xelatex via the Python subprocess.run function. The script uses argument lists instead of shell strings, which is a secure implementation that avoids shell injection risks.
[INDIRECT_PROMPT_INJECTION]: The skill ingests content from external research papers (titles and abstracts) which could potentially contain adversarial text. However, the skill mitigates this surface by providing the agent with strict YAML schemas and structured checklists, which helps enforce boundaries between the paper data and the agent's instructions.
[DATA_EXPOSURE]: The skill processes and saves research data locally within the project directory (e.g., in sources/ and extracted/ folders). It does not attempt to access sensitive system files, environment variables, or user credentials.

literature-review-ml