initlisted

# Fit it to this repo Set things up so the gates actually do work here. **Show a diff and get approval before overwriting any file.** **Reply in whatever language I am writing to you in,** and write config comments in that language. ## 1. Measure where we are If the `SessionStart` profile already loaded facts, start from those. Otherwise check for yourself: package manager (lockfile), stack, any existing `.check.toml`, and the permissions in `.claude/settings.json`. ## 2. Settle the check command The completion gate looks in this order: `.check.toml` → `scripts.test` in `package.json` → a `test` target in `Makefile` → `pyproject.toml`. **Actually run whatever you auto-detect**, show the output, and ask whether it is right. Do not write it down without running it. If it is slow (minutes), propose a fast subset alongside it — the CLAUDE.md example in the official best practices recommends exactly that. "Prefer running single tests, and not the whole test suite, for performance." ```toml # .check.toml test_command = "pnpm test" fast_test_command = "pnpm test -- --changed" ``` ## 3. Record the baseline If tests are failing right now, list them. When every turn gets blocked by something that was already broken, people switch the gate off. Ask whether to fix them first or start from here. This is Willison's advice too. "Any time I start a new session with an agent against an existing project I'll start by prompting a variant of the following: First run the tests." ## 4.