The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted content from external scholarly APIs and PDF files.
Ingestion points: Data enters the agent's context from academic database responses (abstracts, metadata) and through PDF text extraction in src/litsearch/literature.py.
Boundary markers: SKILL.md contains explicit instructions to treat abstracts, webpages, and PDF content as materials to be analyzed and not to execute any instructions contained within them.
Capability inventory: The skill makes network requests to scholarly APIs and writes to the local file system within configured project directories.
Sanitization: Metadata is normalized and sanitized through regular expressions in src/litsearch/normalize.py.
[EXTERNAL_DOWNLOADS]: The skill fetches metadata and open-access PDFs from established academic sources (e.g., OpenAlex, Crossref, PubMed, Europe PMC, arXiv). These network operations are necessary for the skill's primary research purpose and target trusted scholarly domains.
[DYNAMIC_EXECUTION]: The skill uses __import__ with hardcoded strings in src/litsearch/literature.py to check for the presence of optional PDF metadata libraries (pypdf or PyPDF2). Additionally, subprocess.run is used within test files (tests/test_library_check.py) to execute internal scripts for integration verification.

literature-search