What problem does it solve? Shipyard behavior is controlled by a layered JSON configuration spread across user, repo, and local files, and it is hard to know which layer a value came from or why a change did not take effect. This Skill shows every resolved value with its source layer, validates the whole chain, and reloads settings mid-session without a restart. ## Core Features & Use Cases - Show resolved config with provenance: Reports the three layers (user, repo, local), which exist, and where each value resolves from, including derived defaults computed from the repo. - Validate and diagnose failures: Names every configuration error with its key and layer, covering unset required keys, retired SY_* environment variables, model floor refusals, and CLAUDE_CODE_SUBAGENT_MODEL conflicts. - Set values at the right scope: Guides direct JSON edits to the correct layer file (user, repo, or gitignored local), then validates and confirms the change landed. - Reload mid-session: Re-resolves the layer chain so edits apply to the next dispatch, while refusing to silently absorb config changes during an active /sy:ship run. - Use Case: You changed a model tier but agents still run the old model. Run the config skill to see the value is overridden by the gitignored local layer, fix the right file, and reload so the next dispatch picks it up. ## Quick Start Ask the AI to show the current Shipyard configuration and explain which layer each resolved value comes from.