ralph-loop-kiro-specs

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime prompt to kiro-cli causes the agent to read and include outsider-authored free text from the spec directory files (.kiro/specs/<specs_name>/{requirements.md,design.md,tasks.md,progress.md}) and steering files (.kiro/steering/*.md) into the LLM context via the “Load Context” phase; these documents are authored by the operating user’s collaborators/third parties rather than the skill itself.