The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute several shell commands, including grep, find, and git, to scan the project directory for specific patterns. These commands are used for read-only auditing and are scoped to the project's Assets and configuration files.
[DATA_EXFILTRATION]: While the skill reads sensitive configuration files like gradle.properties and project metadata, it does so to verify security settings (such as TLS protocols) and project integrity. There are no network-based commands (e.g., curl, wget) or patterns indicating that this data is transmitted outside the local environment.
[PROMPT_INJECTION]: The skill analyzes external content (project files) and incorporates results into a report. This presents a surface for indirect prompt injection if the audited files contain instructions designed to influence the agent, but the skill itself contains no malicious injection patterns or attempts to bypass safety filters.

review-unity-specific