The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill includes a shell command example that pipes the output of an API request into a Python process for parsing.
Evidence: Found in SKILL.md within the Unpaywall API section, where curl output is piped to python3 -c.
Analysis: This pattern is used to parse JSON data from the Unpaywall service. The data source is a well-known academic repository, and the Python script is a hardcoded, safe parsing routine that does not execute code from the remote payload.
[EXTERNAL_DOWNLOADS]: The skill initiates network requests to several external academic metadata and repository services.
Evidence: curl commands in SKILL.md target api.openalex.org, api.crossref.org, export.arxiv.org, www.nber.org, api.unpaywall.org, and api.semanticscholar.org.
Analysis: These connections target established and reputable academic infrastructure. The skill includes instructions to include contact information via the mailto parameter, which is a standard best practice for accessing these specific services.
[PROMPT_INJECTION]: The skill's primary function involves retrieving and processing untrusted data from external repositories, creating a surface for indirect prompt injection.
Ingestion points: Fetches paper titles, abstracts, and metadata from APIs like OpenAlex, arXiv, and NBER (documented in SKILL.md).
Boundary markers: The skill does not currently instruct the agent to use specific delimiters or to ignore instructions embedded within the retrieved academic text.
Capability inventory: The agent has access to the Bash and Agent tools, which are powerful capabilities if a successful injection were to occur.
Sanitization: No specific sanitization or validation logic is implemented for the data retrieved from external APIs.

academic-paper-search