hotfix

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

• Third-party content exposure detected (medium risk: 0.65). The workflow uses an outsider-authored free-text input via <bug_description>$ARGUMENTS</bug_description> (user-provided/issue-link/error text that the operating user did not author), which is then embedded into the LLM context in the Phase 0/1 prompts (e.g., Task @codebase-review-agent(... Bug: <bug_description>)).