The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external scientific repositories, which could theoretically contain instructions designed to influence the agent.
Ingestion points: scripts/discover_literature.py fetches paper titles, authors, and abstracts from the arXiv, Crossref, and OpenAlex APIs.
Boundary markers: The skill does not currently use explicit markers or delimiters to isolate fetched scientific metadata from the rest of the agent's context.
Capability inventory: The skill has capabilities for network retrieval (discovery APIs) and local file writing for caching results in ~/.cache/literature-discovery.
Sanitization: Content is parsed structurally using JSON, XML, and HTML parsers, but the textual data is not specifically sanitized to remove potential prompt injection sequences.
[DYNAMIC_EXECUTION]: The skill includes a test suite that dynamically loads the main search script to verify its logic.
Evidence: scripts/test_discover_literature.py utilizes importlib.util and exec_module to load and run tests against discover_literature.py. This is standard developer practice for validating skill code.
[EXTERNAL_DOWNLOADS]: The skill communicates with well-known academic data services to retrieve literature metadata.
Evidence: scripts/discover_literature.py performs network requests to official API endpoints for arXiv, Crossref, and OpenAlex to satisfy user search queries.

research-scientific-literature