The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The installation guide instructs users to clone a repository and install plugins from an untrusted GitHub account (github.com/zLanqing/codex-claude-academic-skills). This involves fetching executable skill definitions and scripts from a remote source.
[COMMAND_EXECUTION]: The skill's documentation and troubleshooting sections recommend running local Python scripts included in the package, such as check_text_overflow.py, to validate presentation layouts. It also describes executing MATLAB or Octave commands for data processing, which could lead to unauthorized local execution if the source repository is compromised.
[EXTERNAL_DOWNLOADS]: The skill uses the requests library to perform network lookups on well-known academic services like doi.org and api.crossref.org to retrieve bibliographic data.
[PROMPT_INJECTION]: The skill processes complex external data sources (PDFs and CSV files) to generate research papers and presentations. This creates an indirect prompt injection surface where instructions hidden within academic papers or datasets could influence the agent's output.
Ingestion points: Ingests user-provided files including paper.pdf, data.csv, features.csv, and labels.csv.
Boundary markers: The instructions do not specify the use of delimiters or specific commands to disregard instructions embedded in the ingested text.
Capability inventory: The skill possesses the ability to read and write files (DOCX, PPTX, PNG, PDF, BIB) and perform network GET requests.
Sanitization: No technical sanitization or validation logic is described for the content extracted from external data files.

codex-claude-academic-skills