What problem does it solve? Evolving a repo-local validation contract into a reusable harness system is risky: custom scripts hardcode linters, recursive file walks hit build folders, and local fixes break downstream consumer projects. This Skill provides a structured lifecycle for generalizing a proven steward.yaml contract into a declarative harness and safely testing it across a producer-consumer repository boundary. ## Core Features & Use Cases - Cascading Agent Surface Architecture: Guides a four-layer design (embedded hooks, protocol adapters, orchestrator CLI, AI skill wrapper) that decouples transport from engine logic. - Cross-Repo Remediation Workflow: Defines a seven-step dogfooding process from sibling baseline through dual verification, with a detour stop rule to prevent promoting unproven fixes. - Proof Artifact Requirements: Specifies required producer and consumer artifacts (schema tests, updated steward.yaml, benchmark summaries) before promoting a generalized harness change. - Use Case: After extending a steward CLI with a new declarative lint rule, deploy the local build to a sibling consumer repo, capture adoption friction, fix root causes in the producer, and verify both suites pass. ## Quick Start Ask the AI to generalize the local steward harness and validate it against the sibling consumer repository using the harness-engineering-lifecycle workflow.
