The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests external data from CSV, TSV, and H5AD files (e.g., counts.csv, metadata.csv) which are user-controlled and could theoretically contain malicious instructions.
Ingestion points: Data is loaded via pd.read_csv, load_count_matrix, and anndata in SKILL.md (Step 2).
Boundary markers: There are no explicit boundary markers or instructions to the agent to ignore embedded natural language instructions within the data files.
Capability inventory: The skill's scripts (PyDESeq2, gseapy) perform statistical analysis and print results. No dangerous capabilities like system-wide file writes, privilege escalation, or arbitrary network exfiltration were detected.
Sanitization: The instructions do not specify sanitization for the content of the data files before processing.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of several external Python packages.
Packages: pydeseq2, gseapy, pandas, numpy, scipy, anndata, and tooluniverse are listed for installation via pip.
Vendor Resources: The package tooluniverse is a resource associated with the author 'wu-yc'.
[DYNAMIC_EXECUTION]: The skill provides instructions for using rpy2 to execute R code (specifically clusterProfiler) from within the Python environment.
Execution Method: Uses rpy2.robjects.packages.importr and pandas2ri to bridge Python data to R functions for enrichment analysis. While standard for this workflow, it involves dynamic execution of code in a secondary runtime (R).

tooluniverse-rnaseq-deseq2