The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches data from several well-known academic and scientific service domains, including OpenAlex, NCBI (PubMed), Europe PMC, arXiv, Crossref, and Zenodo. These operations are essential to the skill's primary function of literature retrieval.
[REMOTE_CODE_EXECUTION]: The install.sh script executes curl -LsSf https://astral.sh/uv/install.sh | sh. This pattern downloads and executes the installer for 'uv', a well-known and widely used Python package manager from Astral. As a well-known utility, this is considered a safe installation practice for this skill's ecosystem.
[COMMAND_EXECUTION]: The skill makes several subprocess calls to manage its environment:
Invokes the uv CLI to run the lit search utility.
In src/litsearch/screen_host.py, it invokes the claude CLI via asyncio.create_subprocess_exec to support local model screening when an API key is not provided.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data (titles and abstracts) fetched from external academic APIs and passes them to an LLM for relevance screening in src/litsearch/screen_runner.py.
Ingestion points: Data enters the system via src/litsearch/sources/*.py through various API endpoints.
Boundary markers: Prompts use Markdown headers (e.g., ### key: ...) to delimit records, and specific instructions in src/litsearch/screen.py define the screening logic.
Capability inventory: The skill has capabilities for network access and file writing within the project run directory.
Sanitization: src/litsearch/normalize.py includes a clean_text function that removes HTML tags and collapses whitespace, providing a basic layer of sanitization for ingested metadata.

lit-search