---
name: verify
description: Run the project's verification command and paste its output verbatim.
  Use as the last step of any task, or when asked to prove a change works.
when-to-use: prove it, run the tests, verify, is it green
allowed-tools: shell
user-invocable: true
---

Place at `.grok/skills/verify/SKILL.md`. Invoke as `/verify`, or let the model pick
it up when a brief ends with "verify".

The point of this skill is the last rule. Agents summarise test output because it is
long; the summary is where "3 failures" becomes "mostly passing".

---

1. Find the verification command. In order: the VERIFY line of the current brief;
   the test command in `AGENTS.md`; otherwise detect it (`package.json` → `npm test`,
   `*.sln`/`*.csproj` → `dotnet test --nologo -v q`, `pyproject.toml` →
   `uv run pytest -q`). If none can be found, say so and stop — that is a problem to
   solve before the task, not after.
2. Run it exactly once. Do not modify anything first.
3. Paste the output **verbatim** — the full summary block, including counts, every
   failure name and its message. Do not summarise, paraphrase, or say "tests pass"
   without the output that shows it.
4. State in one line whether the result meets the brief's VERIFY expectation. If it
   does not, stop there: do not start fixing. The fix is a new task with its own brief.