scopelisted
Install: claude install-skill rishabhsisodiya/stop-guessing
## What this skill does
Turns an idea into an ordered, coarse, living plan, and keeps it honest as you ship.
It answers **what to build, in what order, and which parts need a decision first**.
It does not answer *how* to build any one of them. That is `/architect` (the decision)
and `/develop` (the code).
One command, inferred mode. Never a subcommand:
- **plan** (default): no scope yet, or you are planning the next slice.
- **replan**: a scope exists and you ran `/scope` bare. Reports where things stand,
reconciles what shipped, and queues what is next. This is the normal rhythm, not a rare event.
- **add**: a scope exists and the argument names one feature. Enrolls one row without replanning.
## The core rule: this file stays coarse
A scope that lists build tasks rots within a week, because the tasks change every day and
nobody updates two places. So:
- A feature that is **not yet designed** gets **exactly one checkbox**: its entry command.
- Atomic build steps live in the spec's build plan, written by `/architect`. Never here.
- The scope carries intent, order, status, and one `Done when:` line. Nothing else.
## What this skill refuses to do
These refusals are the skill. Read them as hard rules, not preferences.
1. **Never name a tool or make a design choice.** No library, ORM, provider, host or
framework, and no endpoint, table, field, screen layout, or "extend X rather than add Y".
Those are *how*, and how belongs to `/architect`. Writing it here silent