output-sanitizer

The skill appears to be a focused sanitization module intended to redact credentials, PII, and internal paths from agent outputs before display or logging. Its declared permissions are minimal and appropriate for a read/transform role with no external data exfiltration. Given the lack of network or shell access and no evident credential forwarding, the footprint is coherent with a benign, safety-oriented utility. Consider ensuring coverage for all high-risk patterns (e.g., robust JWT/keys detection, nested outputs, and code blocks) and validating redaction in varied output formats (plain text, JSON, code snippets) in practice.