The Agent Skills Directory

[SAFE]: The skill contains instructional content and code templates for extending a specific software framework. The operations described, such as creating reward computation scripts and registering them in an initialization file, are standard development tasks and do not exhibit malicious intent.
[PROMPT_INJECTION]: The reward functions documented in the skill process model completions, which constitutes a potential ingestion point for untrusted data. While this is an inherent part of the skill's purpose, it represents the primary attack surface. Ingestion points: 'completions' parameter in Python reward functions (SKILL.md). Boundary markers: Absent. Capability inventory: Creation and execution of Python scripts within the project directory. Sanitization: Absent.

add-reward