sandcastle

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

• Third-party content exposure detected (high risk: 0.95). Yes—step 1’s backlog resolution can fetch outsider-authored free text (e.g., GitHub issue body and comments via gh issue list ... --json ... body,comments[].body, or Linear ticket descriptions) and then injects that literal text into the implement prompt (the LLM context) via bang-substitution / @-included tickets.md.