The Agent Skills Directory

[DATA_EXFILTRATION]: The skill transmits clinical queries to api.openevidence.com and retrieves medical citations. This network activity is consistent with the skill's primary purpose as an integration for the OpenEvidence platform.
[INDIRECT_PROMPT_INJECTION]: The skill ingests clinical query text which may originate from external clinical environments. It implements mandatory de-identification steps in SKILL.md (deidentifyPHI) and references/implementation.md (sanitizeForOpenEvidence) to redact SSNs, MRNs, and provider names using regex-based filtering before data transmission.
Ingestion points: ClinicalQuery.queryText processed by submitClinicalQuery in SKILL.md.
Boundary markers: The logic employs functional encapsulation to ensure de-identification occurs before network calls.
Capability inventory: Uses fetch for network operations and environment variables for authentication.
Sanitization: Multiple regex patterns are used to identify and redact sensitive identifiers (SSN, MRN, phone, email, DOB).

openevidence-data-handling