The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external VCF data which represents an untrusted ingestion surface.
Ingestion points: VCF files are ingested in Step 2 (input.vcf.gz), Step 4 (clinvar.vcf.gz, GCF_000001405.40.gz), and the Python parsing logic in Step 6.
Boundary markers: None identified; external data is processed directly without explicit delimiters or warnings to ignore embedded instructions.
Capability inventory: The skill uses subprocess.run to execute Java commands and performs file-write operations via shell redirection (>).
Sanitization: No specific sanitization or validation of VCF content is implemented before processing through shell-based tools or Python scripts.
[EXTERNAL_DOWNLOADS]: The skill fetches software and datasets from external sources.
Fetches the SnpEff core package from a dedicated Azure Blob storage container (snpeff.blob.core.windows.net).
Downloads clinical and SNP reference databases from the official NCBI FTP server (ftp.ncbi.nlm.nih.gov).
Installs packages from the Bioconda registry.
[REMOTE_CODE_EXECUTION]: The skill downloads compressed JAR files and proceeds to execute them using the Java Runtime Environment.
Evidence: Step 1 downloads snpEff_latest_core.zip, unzips it, and then Step 2/3 executes java -jar snpEff/snpEff.jar.
[COMMAND_EXECUTION]: The skill makes extensive use of shell commands and Python's subprocess module to orchestrate the pipeline.
Evidence: Use of subprocess.run in the Recipe for 'Extract Protein-Changing Variants to pandas DataFrame' to call java -jar and capture its output.

snpeff-variant-annotation