The Agent Skills Directory

[DATA_EXFILTRATION]: The skill transmits biomedical text provided by the user to external research APIs (UMLS UTS and MeSH). These services are operated by the National Library of Medicine (NLM), a well-known and trusted scientific organization. The skill includes explicit warnings in the documentation and code comments regarding the potential risk of Protected Health Information (PHI) exposure when using external APIs, advising users to de-identify data beforehand.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted biomedical text as input for normalization and mapping, creating a standard attack surface for indirect prompt injection.
Ingestion points: Clinical terms are received via the --term argument or read from files specified by the --input argument in scripts/main.py.
Boundary markers: The skill does not use specific boundary markers or "ignore instructions" delimiters for the processed text.
Capability inventory: The skill performs network read operations (API lookups) and local file writes (saving mapping results to disk).
Sanitization: Input terms are normalized in scripts/main.py using regular expressions to collapse whitespace and remove non-alphanumeric characters before they are used for matching or sent to APIs.

bio-ontology-mapper