The Agent Skills Directory

[PROMPT_INJECTION] (LOW): Indirect Prompt Injection surface detected. The skill ingests metadata (titles, abstracts, authors) from external APIs (CrossRef and Google Scholar) and interpolates it into the agent's context as BibTeX or JSON.
Ingestion points: doi_to_bibtex.py (line 52: response.text), search_google_scholar.py (line 82: scholarly.search_pubs(query)).
Boundary markers: Absent. The skill does not wrap external content in delimiters or provide 'ignore instructions' warnings to the agent.
Capability inventory: Both scripts possess file-write capabilities (-o flag) and network-read capabilities.
Sanitization: Absent. There is no validation or filtering of the retrieved metadata to prevent embedded instructions from influencing the agent's behavior.
[EXTERNAL_DOWNLOADS] (LOW): The script search_google_scholar.py requires the scholarly library, which is an external dependency not included in the standard library. While a known tool, its installation expands the attack surface.
[DATA_EXFILTRATION] (LOW): The scripts perform network GET requests to doi.org and Google Scholar. While these actions are required for the skill's functionality, they target non-whitelisted domains.

citation-management