mdcp-design-architecture

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime workflow instructs the agent to use a user-provided WORK_ITEM and WORK_ITEM_LOOKUP to inspect the repository and “pull only the shards, docs, and code paths needed for this task,” which means outsider-authored free text could be ingested from user-provided ticket/issue content (e.g., arbitrary ticket text) as part of repo-scoping for the design.