# Start: Draft a sampling plan

Use this to turn a sampling purpose, population and frame description into a source-linked design draft. It does not select respondents or establish causal attribution.

**Minimum inputs:** What you need to learn; the indicator/question and intended population/unit; what is known about the sampling frame. Add reporting subgroups, precision or power requirements, timing, mode, resources and permissions if available. Unknowns can remain explicit.

**Code starter:**

> Use /me-workspace:sampling-plan-draft on [explicit files or scoped folder]. Draft sampling-plan.md using the supplied purpose, population, frame and constraints. Separate source facts from proposed choices. Show applicable sizing arithmetic or a calculation specification with missing parameters. Do not invent precision, power, response rates or approvals. Save and inspect the actual Markdown file.

**Browser:** Upload readable attachments and the individual skill package, or paste SKILL.md with the same request using attachment names. Do not paste machine paths. If file creation is unavailable, request editable Markdown and record that no download was produced. The Code route was checked; browser upload was checked on a paid plan for one skill only.

Open examples/input.md for a fictional ordinary case, examples/incomplete.md for conflicting constraints and examples/unsuitable.md for a boundary case. examples/output.md is a manually authored illustration for the ordinary case, not observed model output. An editable outline is in template.md.

Before using a draft, check population/frame coverage, calculation assumptions, subgroup precision, selection feasibility and decisions needing approval. Save original sources, actual draft and a short handover before pausing. The free files do not supply a Claude account or guarantee completion within provider limits.