The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts r_deseq2_wrapper.py, r_edger_limma_wrapper.py, and convert_rds_to_csv.py utilize the subprocess module to run Rscript and R commands. These executions are scoped to scientific data processing and include explicit workspace isolation to prevent unintended file modifications.\n- [DYNAMIC_EXECUTION]: The skill dynamically generates R script files from internal templates at runtime, incorporating parameters such as design formulas and file paths, and subsequently executes them via Rscript. This is a core functional component for interfacing with the R statistical ecosystem.\n- [INDIRECT_PROMPT_INJECTION]: The skill is instructed to prioritize information from executed notebooks (*_executed.ipynb) found in the data directory. Untrusted data enters the agent context via these files (Ingestion points). While the agent can execute commands and platform tools (Capability inventory), the skill does not specify delimiters for the notebook content (Boundary markers) or explicit text sanitization (Sanitization), relying instead on platform-level safety and standard data validation.

tooluniverse-rnaseq-deseq2