The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill is designed to review external code (ingestion point) and has access to powerful tools including Bash, Write, and Edit (capability inventory). This creates an attack surface where instructions embedded in the reviewed code could influence agent behavior. No boundary markers or sanitization logic are defined in the instructions to mitigate this risk.
Command Execution (SAFE): The skill includes a 'Memory Protocol' that executes a bash command (cat) to read a local learning file. This is used for context persistence and does not interact with sensitive system files or network resources.
Data Exposure (SAFE): No hardcoded credentials or access to sensitive user configuration files (e.g., SSH, AWS) were detected.

gitflow