adversarial-tanstack

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill requires composing a fully self-contained reviewer prompt that embeds the target files and also mandates per-rule "evidence" as either file:line or a quote — meaning a blind reviewer LLM will receive file contents (which may contain secrets like env vars) and is allowed/likely to include verbatim quoted snippets (thus exfiltrating secrets) in its output.