---
name: brief
description: Write a complete task brief — goal, context, constraints, verification — before any work starts. Refuses to proceed without a verification step.
agent: agent
argument-hint: one line describing what you want done
---

Place at `.github/prompts/brief.prompt.md` (repository) or in a directory listed in
`chat.promptFilesLocations`. Run it by typing `/brief <one line>` in chat or the CLI.

The point of this file is the last section. Most briefs omit it, and a brief without
a verification step produces work that merely *looks* done.

---

The user wants: **${input:task}**

Before doing anything, produce a brief in exactly this shape. Ask me for anything you
cannot fill in from the request, the repository, or `AGENTS.md` — but do not ask
obvious questions, and do not start work until every section is filled.

```
GOAL        what "done" looks like, in one sentence
CONTEXT     the files involved; the existing pattern to follow (name it);
            the issue or PR this relates to, if any
CONSTRAINTS what must not change; what is out of scope;
            libraries or approaches that are and are not acceptable
VERIFY      the exact command that proves the goal is met, and the
            output you expect from it
```

Rules:

- If I cannot give you a VERIFY line, propose one from the repository's test setup
  (see `AGENTS.md` for the test command). If there is genuinely no way to verify the
  work, say so and stop — that is a problem to solve before the task, not after.
- Prefer a single test file or filter over the whole suite for the iteration loop;
  name the full-suite command separately for the final check.
- Once the brief is complete, restate it, then carry it out. When finished, run the
  VERIFY command and paste its output verbatim. Do not summarise it. Do not say "tests
  pass" without the output that shows it.