The Agent Skills Directory

[DATA_EXFILTRATION]: The script scripts/onekgpd_api.py performs network operations to db.dnaerys.org to fetch genomic variant data. While this is the primary functionality of the skill, the domain is not among the predefined whitelist of common services. No sensitive local files are accessed or transmitted.
[INDIRECT_PROMPT_INJECTION]: The skill ingests genomic annotations and metadata from an external API (db.dnaerys.org) and local assets. This data is processed by the agent, which has access to the Write Bash tool. This creates an attack surface where data from the database could theoretically attempt to influence the agent's behavior.
Ingestion points: scripts/onekgpd_api.py (fetches remote API data) and assets/kgpe.json (processes local metadata).
Boundary markers: The instructions do not provide specific delimiters or warnings to the agent regarding potential instructions embedded within scientific data fields.
Capability inventory: The agent is permitted to use the Write Bash tool, allowing for code generation and execution based on the skill's output.
Sanitization: The scripts return structured JSON, but no natural language sanitization is performed on the descriptive annotation fields (such as ClinVar terms or consequences) before they enter the agent's context.

onekgpd