The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read Unity project configuration files (Packages/manifest.json, ProjectSettings/ProjectVersion.txt), source code, and local PDF-derived documentation. While these represent surfaces where untrusted data could influence the agent, the skill defines a structured optimization workflow and relies on measured evidence, which helps constrain the agent's actions.
Ingestion points: Unity project metadata, source code files, and local documentation files via rg.
Boundary markers: Not specified.
Capability inventory: File system read/write, rg for searching, and usage of the internal unity-cli tool for profiling.
Sanitization: No explicit sanitization of file content is mentioned.
[COMMAND_EXECUTION]: The skill involves executing shell commands via rg (ripgrep) to search documentation and unity-cli to interact with the Unity editor. These tools are used for project analysis and performance monitoring within the workspace environment. The instructions include a requirement to verify the CLI version and identity gate before use.

unity-optimization