portal-dev-setup

The skill’s Docker setup, credentials, and integration steps broadly fit its stated purpose, and the scanner’s command-substitution findings are documentation artifacts. The main risks are unverifiable/unpinned image distribution, transitive skill installation, and the acknowledged lack of secret isolation between containers. No evidence of credential theft, covert exfiltration, or malicious load-time execution appears in the supplied content.

High-severity command-injection risk: `cmd_add` parses externally supplied or locally supplied manifest fields and evaluates them as host-shell code without safe escaping, allowing a crafted manifest to execute commands as the script user. The manifest can come from `--manifest`, `--from`, or configured `MANIFEST_STORE`; no specific malicious domain or URL is hardcoded in the visible fragment. Registry credentials are decoded and passed to Docker login via stdin, with no visible exfiltration. This is a serious security vulnerability but does not by itself establish that the code contains an active malware payload.