The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill uses batch_template_generator.R to generate N individual R scripts by performing string replacements (gsub) on a base template (base_template_knhanes.R). These generated scripts are then executed at runtime using the source() function within a master runner script (run_all.R). While this is the intended core functionality, it represents a dynamic code generation and execution pattern.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from a combinations.csv file to define analysis parameters. The exposure_coding and outcome_coding fields from this CSV are directly interpolated into the generated R scripts' code blocks.
Ingestion points: The batch_template_generator.R script reads the user-provided or database-derived combination matrix file.
Boundary markers: None identified. There are no explicit delimiters or instructions to ignore embedded code within the CSV fields.
Capability inventory: The generated scripts have full access to the R environment, including file system access (read.csv, write.csv) and potential shell execution. The skill is also granted Bash and Write tool access in its configuration.
Sanitization: No sanitization or validation is performed on the strings extracted from the CSV before they are injected into the R template. A malicious input in the exposure_coding field could potentially break out of the intended R mutate() call and execute arbitrary commands.
[COMMAND_EXECUTION]: The skill workflow involves executing shell scripts (run_all.sh) and R scripts that interact with the local file system and potentially other system resources to process large medical datasets.

batch-cohort