minding

SUSPICIOUS. The skill’s behavior is broadly consistent with a personal graph-memory assistant, but it depends on an externally provided `nks_*` MCP toolchain whose provenance and release trail are not verifiable from the provided evidence. That black-box dependency is granted access to broad personal and cross-realm data, including writes and webhook creation, which is disproportionate to trust without stronger source verification. I found no direct credential-theft, obfuscation, or malicious exfiltration instructions in the skill text itself.