About refine
A workflow skill for turning vague entries in `brain/todos.md` into self-contained, actionable prompts. It uses targeted questions to gather missing details, then updates or removes items and reports the changes.
Key features
- 01Reviews open todo items
- 02Identifies missing action details
- 03Groups clarifying questions
- 04Rewrites items in place
- 05Removes completed or irrelevant items
- 06Summarizes changes
Use cases
- 01Clarifying an underspecified bug report
- 02Turning a vague feature request into agent-ready work
Install
Copy the command and run it in a terminal.
$
npx skills add https:/After installing, type/refinein Claude Code to confirm it loaded.
Details
From the project READMERefine
Read brain/todos.md and refine unclear items into actionable prompts through targeted questioning.
Workflow
Use Tasks to track progress. Create a task for each step below (TaskCreate), mark each in_progress when starting and completed when done (TaskUpdate). Check TaskList after each step.
- Read
brain/todos.mdand review items under## Open. - Triage each item: is it clear enough to act on without further context? A clear item specifies what is wrong, where it happens, and what the fix should be. Skip items that already meet this bar.
- Ask clarifying questions about unclear items using
AskUserQuestion. Group related questions — aim for one round of questions, not a back-and-forth (to minimize back-and-forth and user fatigue). Good clarifying questions target:- What — what exactly is the problem or desired behavior?
- Where — which part of the app, which component, which flow?
- How — how should it work instead? What does "fixed" look like?
- Rewrite unclear items in-place using
Edit, incorporating the user's answers. Each rewritten item should be a self-contained prompt that another agent could act on without additional context. - Remove any items the user says are already done or no longer relevant.
- Report a summary of what changed.
Comments
Markdown supported