The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill performs network requests to the NCBI E-utilities API (eutils.ncbi.nlm.nih.gov) to fetch variant data from ClinVar and dbSNP databases. These are well-known scientific services provided by the National Center for Biotechnology Information.
[INDIRECT_PROMPT_INJECTION]: The skill contains a vulnerability surface where untrusted data or metadata could influence agent behavior.
Ingestion points: scripts/main.py processes user-supplied variant identifiers (rsIDs, HGVS, VCF) from the command line or text files. Additionally, the file variant-annotation_audit_result_v1.json contains pre-computed safety verdicts and audit results.
Boundary markers: No explicit instructions or delimiters are used in the processing scripts to instruct the agent to ignore potentially malicious instructions embedded within processed data.
Capability inventory: The skill possesses network access capabilities (urllib.request) and file system access (open() for reading inputs and writing JSON/text outputs).
Sanitization: The _parse_variant_input method in scripts/main.py uses strict regular expressions to validate and filter variant strings, which mitigates the risk of processing non-compliant or malicious input.
Self-referential content: The inclusion of variant-annotation_audit_result_v1.json presenting a 'security: PASS' status represents a self-referential claim designed to target analysis tools or influence agent interpretation of the skill's safety profile.

variant-annotation