The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted binary artifacts from external game archives and compiled shaders, which presents a surface for indirect instructions to be passed to the agent.
Ingestion points: The skill uses tools like inspect_artifact and extract_artifact to read from engine-specific archives such as .pck and .pak files in SKILL.md.
Boundary markers: The instructions lack specific delimiters or negative constraints to prevent the agent from following instructions that might be embedded within the extracted asset metadata or reconstructed code logic.
Capability inventory: The agent has the capability to reconstruct engine scene graphs, game loop logic, and C# MonoBehaviour scripts based on the recovered data.
Sanitization: There are no described mechanisms for sanitizing or validating the extracted binary data before it is interpreted by the agent.

rea-game-asset-recovery