decision-eval

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

• Third-party content exposure detected (high risk: 0.70). The skill reads an ADR/ARCH markdown file from the repository (docs/decisions/ADR-*.md or docs/architecture/ARCH-*.md) and passes the file path to the decision-scorer agent, which will ingest that outsider-authored document text (i.e., content not authored by the operating user) into the LLM context.