The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes user-supplied arguments to derive branch names, which represents a potential surface for indirect prompt injection.
Ingestion points: The $ARGUMENTS variable is used in the SKILL.md file as the source for branch naming.
Boundary markers: External data is encapsulated within <context> tags.
Capability inventory: The skill utilizes git checkout, git status, and a local script detect-base-branch.sh.
Sanitization: The instructions include mandatory kebab-case conversion for branch names and require explicit user approval via the AskUserQuestion tool before any commands are executed, effectively mitigating the risk of malicious input influencing the terminal operations.

create-branch