design-spec-extraction

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill explicitly instructs the agent to extract visible text content from screenshots ("content (text, icons)"), write those values verbatim into intermediate and final JSON files, and even place the full JSON into a ticket — which can force the LLM to output any secret values present in the images.

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

• Third-party content exposure detected (medium risk: 0.30). In SKILL.md the workflow requires extracting UI text/content from an outsider-provided screenshot input (“Screenshot: [attached image]”), and the later passes read and consolidate those extracted results from intermediate files on disk.