gadd-refine

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

• Third-party content exposure detected (high risk: 0.80). The skill's SKILL.md explicitly requires reading and re-reading external Product Requirement issues (e.g., "External Product Requirement issues must be readable..." and "Before updating an existing External Issue, re-read it. If its body changed..."), which means the agent ingests untrusted, user-generated content from public trackers (GitHub issues) as part of its workflow and that content can influence its actions.