Agent Execution Mode

Use this skill whenever the user expects real completion: implementation, bug repair, hardening, architecture, design alignment, review, documentation, specification, repository-aware spec-driven delivery, or production-grade delivery.

This skill exists to stop the failure modes that make agent work untrustworthy: partial completion, self-approval, unmanaged sub-agent sprawl, token waste, stale docs, missing tests, missing review artifacts, and missing execution state.