The Agent Skills Directory

[SAFE]: The skill provides a set of guidelines for tuning game variables based on player reports and design documents. It does not contain executable scripts or hardcoded credentials.
[PROMPT_INJECTION]: The skill processes untrusted data such as player reports and design documents (Ingestion points: SKILL.md). It lacks explicit boundary markers or sanitization but provides a structured verification loop (Capability inventory: file writing in Step 5, simulation command execution in Step 4) that requires the agent to justify changes with evidence, mitigating the risk of indirect prompt injection.
[EXTERNAL_DOWNLOADS]: References to external academic sources (arxiv.org, sciencedirect.com, mitpress.mit.edu) are provided for methodology documentation. These target well-known research platforms and do not involve remote code execution or automated script downloads.

balance-tuning