skill-learning

SUSPICIOUS: the skill’s core behavior is coherent, but it combines untrusted external-content ingestion with local skill modification and a third-party fetch relay. Main risk is indirect prompt injection/persistence into SKILL.md files, not malware or installer abuse.