The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests genomic metadata from an external API (regulomedb.org), which represents a theoretical attack surface for indirect prompt injection. However, given the nature of the source (an established scientific database), the risk is minimal.
Ingestion points: Data is retrieved from the RegulomeDB REST API in functions like regulome_score, score_variant, scan_region, and throughout the example workflows.
Boundary markers: None implemented in the instruction set.
Capability inventory: The skill possesses network read capabilities (requests.get) and local file write capabilities (pandas.to_csv, matplotlib.pyplot.savefig).
Sanitization: The skill performs basic type casting for numeric fields but does not sanitize descriptive strings returned by the API.
[SAFE]: All network requests are directed toward the official scientific domain for RegulomeDB (regulomedb.org), which is a well-known resource maintained by the ENCODE project.
[SAFE]: The skill depends on standard, reputable Python packages for data science (requests, pandas, matplotlib).
[SAFE]: No malicious patterns such as credential harvesting, obfuscation, privilege escalation, or persistence mechanisms were detected in the instructions or scripts.

regulomedb-database