The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides an interface to query and ingest data from an external clinical evidence API, which introduces a surface for indirect prompt injection from untrusted medical literature content.
Ingestion points: Data enters the agent's context through the EvidenceResponse and Citation objects retrieved from api.openevidence.com (SKILL.md).
Boundary markers: The implementation patterns do not demonstrate the use of delimiters or specific instructions to the agent to treat the fetched clinical data as potentially untrusted content.
Capability inventory: The skill requires Read, Write, and Edit permissions, and includes code for making network requests and transforming API responses (SKILL.md, implementation.md).
Sanitization: The code samples provide structure transformation but lack explicit sanitization or filtering of the textual content (such as abstract or answer fields) returned by the remote API.
[SAFE]: The skill uses appropriate secret management and targets legitimate service endpoints.
Secrets: The client correctly retrieves credentials from the environment variable OPENEVIDENCE_API_KEY rather than hardcoding them (SKILL.md).
Network: All network operations are directed at the official API of the service the skill is designed to support (https://api.openevidence.com).

openevidence-sdk-patterns