overnight-parallel-agent-prompts

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

• Third-party content exposure detected (high risk: 0.85). High: the workflow fetches open GitHub issues via gh issue list ... --json number,title,labels and then (in later steps) requires “re-read the latest issue body/comments” and “capture the committed diff … and include issue context, changed files, verification commands/results, and the exact diff” into the LLM prompt—issue bodies/comments are outsider-authored free text that can contain indirect prompt-injection payloads.