judgelisted

<!-- Generated by scripts/build-client-adapters.py from clients/workflows.json; schema v1; client github-copilot-cli. --> # Judge a staged diff Select `adr-kit:judge` from `/skills`, or ask Copilot to use the `judge` ADR Kit skill. Resolve `<plugin-root>` from this installed skill. Use only bundled, local ADR Kit tools and follow this canonical workflow: 1. Read the staged diff and stop clearly when it is empty. 2. Call `adr-kit.adr_judge`; for each violation offer supersession, a narrow new decision, or the smallest compliant code change. 3. When the diff implements a linked Proposed ADR or exposes an undocumented consequential decision, offer the exact client-native grill invocation without weakening Accepted ADR enforcement. 4. Evaluate `llm_judge: true` rules in the active session and never send code to another provider without permission. 5. On a settings request, run `python <plugin-root>/bin/adr-judge --adr-dir docs/adr --show-config` and report every value with its provenance, including an `available: false` backend rather than summarizing it away. 6. Change the judge backend only through `adr-judge --set-backend host --host-client <this client's id>` (the host model is the only backend, ADR-036); never hand-edit .adr-kit.json, never ask for an API key, and never write a credential to a file, because that file is committed and the judge refuses a key found in it. Do not contact another model or mutate user-owned instructions.