The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: An automated scanner flagged a command piping curl output to python3. Analysis confirms this is a false positive: the skill pipes scientific metadata from the OpenAlex API into a local citation-processing script (scripts/openalex_to_bibtex.py) rather than executing remote content as code.
[EXTERNAL_DOWNLOADS]: The skill makes network requests to api.openalex.org to retrieve paper metadata and abstracts. OpenAlex is a well-known and trusted academic indexing service.
[COMMAND_EXECUTION]: The skill utilizes several system utilities to manage the document lifecycle, including bash, mkdir, cp, sed, find, and curl. It also invokes the pdflatex and bibtex compilers to produce the final PDF document.
[PROMPT_INJECTION]: The skill is susceptible to 'Indirect Prompt Injection' through its ingestion of untrusted scientific data.
Ingestion points: Scientific data is retrieved from the OpenAlex API via curl and processed by the skill's Python scripts.
Boundary markers: No delimiters or isolation techniques are used to separate the external text from the LaTeX document structure when writing to .tex or .bib files.
Capability inventory: The skill utilizes pdflatex, which has the capability to read and write files on the host system. The workflow also performs multiple shell-based file manipulations on the generated source code.
Sanitization: The scripts/openalex_to_bibtex.py script performs ASCII normalization and escapes some special characters (e.g., &, %, #, _). However, it does not comprehensively escape LaTeX control characters such as backslashes (\\) and curly braces ({, }), which could allow a malicious database entry to inject commands that are executed during the compilation phase.

paper-creator