agentmemory-persistent-memory

SUSPICIOUS: the skill’s core behavior mostly matches its stated purpose, but install trust is materially weaker than the documentation suggests. The main concern is supply chain: a package from an unrelated maintainer is executed via npm/npx, with upstream docs indicating a secondary iii engine install outside the primary package boundary, plus unpinned @latest examples. Data collection is broad but proportionate to a persistent-memory tool, and I found no direct credential theft, hidden exfiltration, or malicious pre-execution behavior.