The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches and processes scholarly metadata, including titles and abstracts, from the OpenAlex API. This external data is untrusted and could contain malicious instructions designed to influence the agent's screening or annotation behavior.
Ingestion points: Untrusted data enters the context via API responses in phases/phase1-search.md and phases/phase3-snowball.md.
Boundary markers: The skill does not provide explicit delimiters or instructions to the agent to ignore potentially malicious content embedded in the abstracts.
Capability inventory: The skill allows for Python execution (network requests, file I/O) and shell command execution across its various phases.
Sanitization: No sanitization or validation of the retrieved text is performed before processing.
[DYNAMIC_EXECUTION]: The skill provides Python code templates in several files (e.g., phases/phase1-search.md, phases/phase4-fulltext.md) that the agent is expected to populate and execute to interact with external APIs and manage local data. This involves generating and running scripts from known templates.
[COMMAND_EXECUTION]: In phases/phase6-synthesis.md, the instructions prompt the agent to execute a local shell command (python ingest.py) to process bibliography and PDF files as part of the synthesis workflow.
[EXTERNAL_DOWNLOADS]: The skill performs legitimate network operations to fetch metadata from api.openalex.org and api.unpaywall.org. It also includes logic to download PDF files from URLs provided by these scholarly services.

lit-search