task-board-init

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

• Third-party content exposure detected (high risk: 0.70). The required runtime workflow reads insider-authored free text from the target repo itself—e.g., the Step 1 “Explore” subagent returns DOCS inventory and the Step 4c doc sweep reads legacy backlog/feature/task doc text—so that repo’s existing prose (potentially containing outsider-authored content such as issue/wiki/forum text) can be ingested into the LLM context via the subagents’ prompts/outputs; only the file writes are scoped to .claude/features/**, not the reads.