agent-builder

The skill demonstrates coherent alignment between its stated goal (building and automating Claude Code agents via SDK/CLI without API keys) and its described capabilities. However, there are notable security concerns around the download-and-execute installer pattern, potential credential exposure via ~/.claude/, and a typographical issue in an installation command that could lead to misconfiguration. The data flows are generally contained within the Claude Code ecosystem and local config, which is appropriate for the stated purpose, but the supply-chain risk from remote installers and any external doc fetches warrants caution. Overall, the risk is suspicious due to multiple supply-chain indicators and unpinned installers, but not conclusively malicious given the legitimate intent and controlled tool paths. Treat as SUSPICIOUS with emphasis on securing install sources, pinning versions, and avoiding untrusted shells.