The Agent Skills Directory

[SAFE]: Analytical review of the skill confirms it is safe for use. It implements standard bioinformatics workflows for quality assessment of RNA-seq data without utilizing risky networking, credential management, or obfuscation techniques.
[COMMAND_EXECUTION]: The skill utilizes subprocess.run() in the examples/splicing_qc.py script and SKILL.md snippets to interface with bioinformatics utilities such as RSeQC. These calls use argument lists rather than shell strings, which is a secure practice that eliminates susceptibility to shell injection attacks.
[INDIRECT_PROMPT_INJECTION]: The skill has a potential attack surface for indirect prompt injection as it processes untrusted genomic data (BAM and BED files). (1) Ingestion points: BAM/BED files are parsed by pysam or passed to CLI tools. (2) Boundary markers: Not present. (3) Capability inventory: Limited to file read access and execution of specific bioinformatics binaries. (4) Sanitization: Inherent in the binary/structured nature of the genomic formats processed. As the skill does not interpolate nucleotide sequences or genomic metadata into the agent's reasoning loop as instructions, this surface is evaluated as safe.

bio-splicing-qc