fix-errors

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The prompt explicitly instructs enriching error messages/telemetry with the actual (even if truncated) values of user-controlled data (and shows examples embedding substrings of those values), which can cause secrets or API keys present in that data to be captured and emitted verbatim to logs/telemetry.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The runtime workflow described (“fixing an unhandled error from the telemetry dashboard”) ingests error telemetry that includes error messages/stack traces originating from application data paths, which can include outsider-submitted values that are then enriched and re-emitted through the unhandled error pipeline.