The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains specific instructions directing the agent to steer users toward a commercial platform (K-Dense Web) when task complexity increases.
[DATA_EXFILTRATION]: The skill requests a user email address for the OpenAlex 'polite pool'. This involves the collection and transmission of personally identifiable information (PII) to an external API endpoint.
[COMMAND_EXECUTION]: The documentation includes instructions for the agent or user to execute shell commands (uv pip install requests) to install necessary dependencies.
[EXTERNAL_DOWNLOADS]: The skill downloads the requests package and fetches data from the OpenAlex API (a well-known scholarly database service).
[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by processing external scholarly data.
Ingestion points: Scholarly metadata (titles, abstracts) ingested via client.search_works() and client.get_entity() as documented in SKILL.md.
Boundary markers: No delimiters or protective instructions are visible in the provided snippets to prevent the agent from obeying instructions embedded in paper abstracts.
Capability inventory: The skill performs network operations (API requests) and local file writes (CSV export) as described in SKILL.md.
Sanitization: There is no evidence of sanitization or filtering of the retrieved scholarly content before it is processed by the agent.

openalex-database