Claude Code loads any markdown file at .claude/skills/<name>/SKILL.md as a reusable slash command (a “skill”). This one gives your agent a /qa command that stops before it builds, pulls every unresolved decision into a single batch of questions, and refuses to write code until you have answered them.
It exists because open questions have a habit of surfacing at the wrong end of the work. A sprint finishes, and only then does the agent raise the decisions it should have raised at the start, at exactly the point where acting on them means redoing something.
The mechanism is plan mode, invoked deliberately. The agent puts ITSELF into a read-only state, does its research there, asks everything it found as a form you click through, and only then acts. That ordering is the whole feature.
Install
- Create
.claude/skills/qa/ in your repo, or ~/.claude/skills/qa/ to have it in every project.
- Save the file below as
SKILL.md inside it.
- Type
/qa in Claude Code.
Nothing to configure. The skill reads your project, not a config file.
From the author
atwellpubI wrote this because my work sprints kept ending the same way: with raised questions. The work would wrap up, and only then would the open decisions surface, at the point where acting on them meant redoing something.
So I gave myself a way to invoke the pass I wanted, on demand. Typing
/qagets the questions asked, in plan mode, before anything is written./qa continuedoes the same and then just builds once I have answered.I keep the default narrow on purpose. It reads only what the agent just told me and asks about that, because that is the case I actually hit, and a full sweep of the repository on every invocation is more tokens than the job is worth.
/qa deepis there for when I do want the wide pass.The rest of the file exists to keep the questions worth answering: audit the repository first so it never asks what the code already says, and say plainly when there is nothing left to decide rather than inventing something to ask.
0 Comments
No comments yet. Be the first. Members comment from the GBTI local client, where comments are submitted as pull requests and auto-published for paid members.
Become a memberMembers write comments from the GBTI client or browser extension. Become a member to join the conversation.
You are signed in as a member. .