write-prose-fix

SUSPICIOUS. The skill’s purpose is coherent, but it relies on a load-time pre-execution shell command with unquoted arguments and an unseen local script, creating meaningful command-execution risk disproportionate to a prose-fixing skill. No evidence of credential theft, remote download, or third-party exfiltration appears in the supplied content.

The script's intended purpose is gathering prose-lint context. The primary security concern is execution of the caller-supplied lint string through `eval`; this permits arbitrary shell commands if that input is untrusted. Target paths are also not confined to the repository, allowing unintended file reads. No specific malicious payload or exfiltration behavior is shown in the fragment.