The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run shell commands using file paths derived from the repository. Maliciously crafted file names could lead to command injection if the agent interpolates them into the shell without proper escaping.
[PROMPT_INJECTION]: The skill provides a surface for indirect prompt injection by reading instructions from files like .ai-rules and CODE_REVIEW.md to drive a subagent.
Ingestion points: Files modified in the git history and configuration files such as .ai-rules and CODE_REVIEW.md (via SKILL.md).
Boundary markers: None; there are no instructions to the subagent to distinguish between rules and potentially malicious data embedded in the rules.
Capability inventory: Includes shell command execution through just and the ability to delegate tasks to subagents.
Sanitization: The skill lacks logic to sanitize file paths or validate the integrity of the rule files before processing them.

quality-check