metaclaw-evolving-agent

SUSPICIOUS. The skill's behavior is mostly consistent with its stated purpose, but that purpose is inherently high-trust: it intercepts live conversations, persists them, reuses them to alter future prompts, and may forward them to external RL services. No clear malware or covert exfiltration is shown, but the data collection, prompt-injection feedback loop, third-party backend forwarding, and network-exposed default make this a high-risk skill.