The Agent Skills Directory

[SAFE]: The skill documents a legitimate scientific pipeline for computing bacterial pan-genomes. It uses established bioinformatics tools including Roary, BLAST+, CD-HIT, and MCL, and standard Python data science libraries such as pandas and matplotlib.
[EXTERNAL_DOWNLOADS]: The skill references software installation via mamba and pip from trusted repositories (bioconda, conda-forge, and PyPI). It also points to a visualization script hosted on the official Roary GitHub repository (sanger-pathogens/Roary). These references are documented neutrally as they originate from well-known and established sources.
[INDIRECT_PROMPT_INJECTION]: The workflow involves ingesting external biological data (GFF3 files) and strain metadata. While this constitutes an attack surface, the structured nature of the data and the use of standard parsing libraries (pandas, biopython) significantly mitigate the risk of data-driven instruction injection. No high-risk capabilities (like network exfiltration of sensitive files or shell interpolation of untrusted strings) were identified.
[COMMAND_EXECUTION]: Shell commands provided for installing dependencies and running the Roary pipeline are standard practice for this domain and do not exhibit signs of malicious intent or obfuscation.

roary-pangenome