The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents methods for ingesting and processing various external data formats which can serve as an attack surface for instructions hidden in data files.
Ingestion points: The skill uses libraries like ldtkgo, go-tiled, and aseplayer to load level and animation data from external files as described in resources/SKILL.md and its references.
Boundary markers: There are no instructions or patterns provided to delimit external content or warn the agent to ignore embedded instructions within these data formats.
Capability inventory: The skill documentation includes capabilities for shell execution (build processes in deployment/SKILL.md) and file system writes (save data via gdata in resources/references/procedural-saves.md).
Sanitization: No sanitization or validation logic is specified for metadata or custom fields within the external asset files.
[PRIVILEGE_ESCALATION]: The deployment/SKILL.md file contains instructions for the user to execute sudo commands for installing system-level dependencies on Linux, such as libgl1-mesa-dev and libxcursor-dev.
[COMMAND_EXECUTION]: The skill references several command-line operations for development and deployment, including go build for multiple target architectures, go install for ecosystem tools like wasmserve, and the use of upx for binary compression.

ebiten