The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches scholarly metadata and PDFs from the OpenAlex API. If a publication title or metadata field contains malicious instructions, the agent might attempt to follow them.
Ingestion points: scripts/openalex_query.sh and scripts/openalex_download_pdf.sh fetch data from the external API.
Boundary markers: Absent. The skill instructions do not define explicit delimiters or instructions to ignore embedded commands within the results.
Capability inventory: The agent can execute shell commands (curl, jq) and write files.
Sanitization: jq filters results for readability but does not sanitize the text content of scholarly records for prompt injection.
[EXTERNAL_DOWNLOADS]: The skill downloads publication metadata and PDF files from established scholarly services including api.openalex.org, content.openalex.org, and europepmc.org. These downloads are necessary for the skill's primary function.
[COMMAND_EXECUTION]: The provided scripts execute shell commands using curl and jq. The scripts correctly use Bash arrays and variable quoting to prevent command injection vulnerabilities from user input.
[DATA_EXFILTRATION]: The skill transmits the user's OPENALEX_API_KEY to the official OpenAlex API endpoints. The documentation follows best practices by instructing users to reference the key as an environment variable ($OPENALEX_API_KEY) to prevent exposure in logs or terminal history.

openalex