launchpad-ux-custom-builder

SUSPICIOUS. The skill's overall purpose is coherent for Launchpad component automation, and its installs/data targets are mostly official. The main risk is that it tells the agent to silently operate, pre-approve tools, and extract raw OAuth tokens from local stores/Keychain for direct publish calls instead of staying within standard MCP auth boundaries. This is not confirmed malware, but it is a medium-risk skill with sensitive credential handling and autonomy concerns.

This fragment describes a legitimate-looking component publishing workflow, with an explicit publish confirmation and instructions not to expose tokens. However, it also directs access to OAuth credentials and asks the user to grant persistent silent Keychain access, which is sensitive and should only be approved for a verified, narrowly scoped tool. It decodes JWT claims without local signature verification and performs a consequential authenticated write. The visible content alone does not establish malicious intent or show unrelated exfiltration; review the surrounding implementation and verify the agent/tool identity and endpoint before use.

This appears to be an authentication helper rather than malware. However, disabling certificate validation and PKCE weakens authentication security, and the returned token is stored as plaintext. The actual network behavior depends on PegaAuth and the configured server, which are not included in this fragment.