moai-alfred-practices

This package is a documentation/skill describing agent workflows and tooling for moai-adk. It does not contain explicit malicious code or payloads, but its allowed-tools and recommended practices create a medium security risk: an agent with these permissions can read local secrets, cache them in Task contexts, run shell commands, and send data to external endpoints or perform repository-changing operations. Recommend the maintainers restrict allowed-tools to least privilege, add deny-lists for sensitive filesystem paths, enforce per-action user confirmations for Bash/WebFetch/automation actions, implement domain allowlists for WebFetch, and add TTL/eviction or encryption for cached Task contexts. With those controls the operational risk would be significantly reduced.