The Agent Skills Directory

[PROMPT_INJECTION] (LOW): Indirect Prompt Injection vulnerability surface. The skill processes data from external scholarly APIs and search results which could be manipulated by an attacker to include instructions for the AI agent.
Ingestion points: doi_to_bibtex.py (via CrossRef API) and search_google_scholar.py (via Google Scholar results).
Boundary markers: Absent; the scripts do not use specific delimiters to isolate external data from the agent's instructions.
Capability inventory: The analyzed scripts do not contain high-risk capabilities like arbitrary command execution or local file writing.
Sanitization: Absent; the content is formatted as BibTeX without escaping or validation against injection patterns.
[DATA_EXFILTRATION] (LOW): The skill performs network operations to non-whitelisted domains.
Evidence: requests.get calls in doi_to_bibtex.py to doi.org and network access performed by the scholarly library.
[EXTERNAL_DOWNLOADS] (LOW): The skill relies on external Python packages not included in the pre-approved trusted lists.
Evidence: Dependencies on requests and scholarly packages.

citation-management