The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes data from external sources that could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: The doi command-line argument in scripts/scite_client.py and the JSON response body retrieved from the api.scite.ai endpoint.
Boundary markers: None. The implementation lacks explicit delimiters or instructions to treat external data as untrusted content.
Capability inventory: The skill possesses network access via the requests library to fetch data from remote servers.
Sanitization: The input DOI is interpolated directly into a URL string without validation, and the API response is printed to the agent context without escaping or filtering.
[METADATA_POISONING]: The skill includes an auxiliary file scite-database_audit_result_v1.json that provides self-reported security evaluations.
Evidence: The file contains numerous safety claims, such as "security": "PASS" and "scientific_integrity": "PASS".
Impact: These claims are treated as data to be evaluated rather than authoritative conclusions. Their inclusion is documented as a potential vector for misleading safety analysis results.
[EXTERNAL_DOWNLOADS]: The skill retrieves citation metrics from a remote service.
Evidence: The scite_client.py script makes HTTP GET requests to https://api.scite.ai using the requests library.
Status: The download targets a well-known academic citation analysis service. This operation is documented neutrally as it aligns with the skill's primary stated purpose.

scite-database