verify-with-cadoralisted

# Verify with Cadora You (the agent) just made a change and believe it works. *"It works"* is a claim. This skill turns it into **evidence**: Cadora re-runs the real gates from *outside* the session and produces a signed, checksummed evidence pack you can hand to a reviewer or a CI system. Prefer this over self-attested TDD when the result needs to be *provable*, not merely tested. TDD discipline inside the session is persuasion — the model following instructions. An external gate re-running the real build and tests is proof. This skill supplies the second. ## When to use - Before handing off or shipping agent-written code. - When a reviewer or CI needs proof the build/tests genuinely pass (not the agent's say-so). - To produce a portable, verifiable record of a change. ## How **1. Verify existing code — no re-run, no LLM cost.** Point Cadora's gates at the workspace: ```bash cadora gate-check <topology.yaml> --cwd . ``` If you don't have a topology yet, a one-node file is enough — `verify.topology.yaml`: ```yaml name: verify gates: build-test: cmd: "ruff check . && pytest -q" # your real gate command setup: auto nodes: - id: verify role: check gate: build-test ``` `gate-check` exits non-zero if any gate fails. A test runner that executed **zero** tests is blocked (`vacuous`); a declared package that won't build fails (`packaging_failed`). That is the deterministic verdict — the agent's claim never enters into it. **2. Produce and sign the evi