architecture-compass

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

• Third-party content exposure detected (high risk: 0.75). The required workflow explicitly inspects and ingests the target repository’s ADRs/docs/agent instructions/examples (outsider-authored relative to the skill’s bundled assets) into the LLM context via the “Inspect … repository evidence” and “Read target ADRs and repo rules” steps, creating an indirect prompt-injection risk from repository text.