The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the local Unity project environment (package IDs, versions, and dependency trees), which creates an indirect prompt injection surface if a project manifest contains malicious metadata.
Ingestion points: package_list, package_check, package_search, package_get_dependencies, and package_get_versions in SKILL.md.
Boundary markers: None present in the current instruction set.
Capability inventory: The skill can modify the project state via package_install and package_remove as defined in SKILL.md.
Sanitization: No specific sanitization or filtering of external package metadata is described.
[COMMAND_EXECUTION]: The skill provides tools to install (package_install) and remove (package_remove) software packages within the Unity environment. While these are powerful operations, they are the primary intended purpose of the skill and are accompanied by explicit guardrail instructions regarding execution modes (Approval vs. Bypass) to mitigate risk.

unity-package