What problem does it solve? AI agent orchestration breaks down when agents return vague or fabricated status reports, lose track of findings between rounds, or route work to the wrong specialist. This Skill enforces a strict, machine-readable result envelope so every agent invocation ends with verifiable provenance, complete finding accounting, and correct next-owner routing. ## Core Features & Use Cases - Terminal and continuation envelopes: Emit RESULT-ENVELOPE blocks for completed or blocked work, and CONTINUATION-ENVELOPE blocks for read-only advisory surfaces like recap, status, and handoff. - Canonical outcome vocabulary: Restricts outcomes to completed_owned, completed_diagnostic, route_required, or blocked, with legacy done_with_concerns explicitly rejected for new envelopes. - Finding accounting and learning disposition: Requires every finding to appear in addressedFindings, unresolvedFindings, or observations, and records whether a run produced a captured, deferred, or not-applicable lesson. - Use Case: After a Bubbles audit agent finishes a diagnostic pass, use this Skill to emit an envelope listing unresolved findings with suggested owners so the orchestrator can route each one to the correct specialist. ## Quick Start Compose a valid Bubbles result envelope for the agent invocation that just finished, including outcome, findings, and next-owner routing.
