The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill handles various genomic file formats which can serve as vectors for indirect prompt injection if the files contain malicious payloads intended to manipulate the agent's logic.\n
Ingestion points: The skill ingests untrusted data from multiple genomic file formats including SAM/BAM/CRAM, VCF/BCF, FASTA/FASTQ, and BED/GTF/GFF via classes like pysam.AlignmentFile, pysam.VariantFile, pysam.FastaFile, pysam.FastxFile, and pysam.TabixFile (documented in SKILL.md, references/variant_files.md, and references/sequence_files.md).\n
Boundary markers: No boundary markers or delimiting instructions are provided to help the agent distinguish between data and potential instructions within the processed files.\n
Capability inventory: The skill allows for the execution of external samtools and bcftools commands using pysam.samtools and pysam.bcftools. It also provides methods for writing files, such as AlignmentFile.write(), VariantFile.write(), and standard Python file operations in workflow examples (documented in SKILL.md and references/common_workflows.md).\n
Sanitization: There is no evidence of input validation or sanitization for the content read from these genomic files before it is processed or used in operations.\n- [COMMAND_EXECUTION]: The skill facilitates the execution of samtools and bcftools through the pysam Python interface.\n
Evidence: SKILL.md and references/alignment_files.md detail how to use pysam.samtools for tasks such as sorting, indexing, and viewing genomic files, which involves spawning subprocesses to run these command-line utilities.\n- [DYNAMIC_EXECUTION]: The skill uses the pysam library to wrap functional calls to external bioinformatics binaries.\n
Evidence: Usage of pysam.samtools and pysam.bcftools to perform operations like sort, index, and view involves runtime execution of external code based on the parameters provided to the Python functions.

alterlab-pysam