gds-ux

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

• Third-party content exposure detected (high risk: 0.70). SKILL.md’s required runtime workflow includes “Source scan” and “Brain dump first” that explicitly instruct subagent-extracts on developer-confirmed planning artifacts (GDD/PRD/narrative) and then later “Subagent reads … sources” for distillation, so any outsider-authored free text in those artifacts can be ingested into the LLM context via the file-reading/extraction path.