meta-prompting

SUSPICIOUS: the core prompt-iteration behavior is coherent and mostly benign, but the footprint expands to unverifiable local components and a custom Python engine that may receive API keys. The skill is not overtly malicious and contains no pre-execution payloads or clear exfiltration, yet its undocumented dependency chain and credential-forwarding example make it higher risk than a pure prompt-engineering skill should be.