The Agent Skills Directory

[SAFE]: The skill provides an educational and functional example of how to implement cost-saving routing logic in AI agents. No malicious behaviors or security vulnerabilities were detected.
[REMOTE_CODE_EXECUTION]: The provided Python code uses the well-known 'openai' library for standard API interaction. There are no attempts to execute arbitrary shell commands or download remote scripts.
[DATA_EXFILTRATION]: The skill does not access sensitive local files (like SSH keys or AWS credentials) or perform unauthorized network requests. All network activity is directed to standard LLM provider endpoints.
[PROMPT_INJECTION]: The instructions do not contain commands to ignore safety guidelines, reveal system prompts, or bypass ethical constraints. The usage of the term 'MANDATORY' is confined to architectural implementation advice for cost management.
[OBFUSCATION]: The code and documentation are written in clear, plain text with no use of Base64 encoding, zero-width characters, or homoglyph attacks.
[INDIRECT_PROMPT_INJECTION]: While the skill processes user prompts to classify them, it does not involve file system writes or tool executions that would create an exploitable attack surface. The complexity classification is performed via local string matching heuristics.

model-tier-routing