migrate-config-to-opus-5

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill requires subagents to read every file and "report back per-file structured findings" including the "quoted current text" (exact file excerpts), which forces the model to output any embedded secrets (API keys, tokens, passwords) verbatim.

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

• Third-party content exposure detected (medium risk: 0.30). Step 2 (“Audit itself is read-every-file … Deploy subagents … Give each subagent … the full text of references/audit-patterns.md … read each file completely”) means the runtime ingests free text from the user-selected local files (e.g., ~/.claude or the project directory) that may contain outsider-authored content, without first selecting specific items beyond choosing the surface.