The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill processes external R scripts that could contain malicious instructions designed to influence the behavior of the r-reviewer agent.
Ingestion points: Reads files from the local filesystem using Read, Grep, and Glob tools based on the provided $ARGUMENTS in SKILL.md.
Boundary markers: Absent. There are no instructions to use delimiters or specific safety wrappers when passing code content to the reviewer agent.
Capability inventory: The skill uses Write to save reports and Task to launch subordinate agents. A successful injection could potentially lead to unauthorized file writes or task manipulation.
Sanitization: Absent. The skill reads the raw content of R scripts and passes them directly to the next stage of the workflow without filtering or escaping.

review-r