Plan directory

.agents/skills/plan-directory/SKILL.md — 148 lines of instruction that ship inside the starter. This page is generated from it, so it says what the skill does without republishing it.

When an agent picks this up

The skill description, verbatim — this is the text an agent matches your request against:

Turn a niche idea into a ZeroDirs directory plan — a candidates TSV of real product names and URLs, plus proposed categories and tags blocks for site.config.ts. Covers picking the niche, sizing the taxonomy against the pSEO and pagination thresholds, and the one rule that makes the rest of the pipeline safe: this stage emits names and URLs only, never invented products. Use when asked to plan a new directory, choose categories for a niche, or build a candidate list before generating listings.

What it walks through

1. The one rule
2. Step 1 — pick the niche and the shape
3. Step 2 — size it against the thresholds
4. Step 3 — write candidates.tsv
5. Step 4 — propose the config blocks
6. Hand off
   • how each candidate was sourced, and how many rows you are not sure about

Commands it runs

node seed/check-dataset.mjs <dir> --profile buyer
node seed/check-links.mjs seed/.out/candidates.tsv --json seed/.out/links.json

Files it touches

site.config.ts · seed/check-links.mjs · seed/check-dataset.mjs

Using it

Skills load themselves — you describe the outcome, not the skill. “Add a Compliance category with an icon and seed it” is enough for an agent that can see .agents/skills/; Claude Code reads .claude/skills/, which pnpm skills:sync mirrors from .agents/ on every install. See Working with an agent.