claude-code-agent-architecture

SUSPICIOUS. The skill's stated purpose as a research/documentation guide mostly matches its content, and most scanner hits are documentation artifacts. The main risk is install trust: it asks users to clone a personal reverse-engineering repo and run package installation, which is broader than strictly needed for reading docs and carries medium supply-chain risk, but there is no clear credential harvesting, covert exfiltration, or malicious execution path in the skill text itself.