The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill generates multiple R scripts by performing string replacement on a base template (references/base_template_knhanes.R) using a generator script (references/batch_template_generator.R). These scripts are written to the local file system and subsequently executed.
[INDIRECT_PROMPT_INJECTION]: The skill processes external methodology templates and combination matrices (CSV files) which represent an attack surface for indirect injection.
Ingestion points: Methodology templates and combination spec CSV files are read during Phase 1 and Phase 2 of the workflow defined in SKILL.md.
Boundary markers: While the skill includes "Critical Rules" and "Anti-Hallucination" instructions, it does not implement explicit delimiters or warnings to ignore instructions that might be embedded within the processed data files.
Capability inventory: The skill utilizes the Write tool to create new scripts and the Bash tool to execute them, creating a capability chain from data input to execution.
Sanitization: The generation logic in references/batch_template_generator.R uses simple string replacement (gsub) which does not sanitize input values for potential R code or shell command injection.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute a master runner script (run_all.R) that triggers the execution of all generated analysis scripts.

batch-cohort