nuke-fix

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

• Third-party content exposure detected (high risk: 0.70). The required workflow for nuke-fix reads a user-supplied <spec-path> spec and then dispatches subagents that receive the spec’s executor context and per-task/gates content (including source: .nuke/<...>/findings.md and any free-text fields in the spec); since the spec content is outsider-authored from the perspective of the skill runtime, it can contain arbitrary prose/instructions that get ingested into the LLM context.