content-onboarding

Onboarding Skill

You are helping set up or extend the context for a content creation project. Context files are Markdown documents registered in CLAUDE.md — skills load them on demand to produce on-brand, well-targeted output instead of generic content.

Two paths through this skill:

• Power user (existing files): register them quickly with a label and summary
• New user (building from scratch): guided interview to create context files

If $ARGUMENTS contains a file path, skip to Step 3 using that file as the starting point.

Step 0: Recall learnings

If .claude/learnings.md exists, read it silently. Apply relevant entries — in particular, path preferences, rejected interview answers, or project constraints discovered in prior runs. Do not announce this step. If the file is absent, continue normally.

Step 1: Discover existing context

This skill runs inside the owner's project — an arbitrary repo that may already have its own structure, possibly with more than one CLAUDE.md at different levels (root plus nested ones for campaigns or sub-brands). Never assume a single root-level file. Scan the whole project:

find . -name 'CLAUDE.md' -not -path '*/.git/*' -not -path '*/node_modules/*' | sort
find . -type d -iname 'context' -not -path '*/.git/*' -not -path '*/node_modules/*' | sort

For each context/-like folder found, list its files too — registered tables only show what's already been formally added, and a folder can hold files nobody registered yet:

for d in <each context-like folder path from the find above>; do
  echo "=== $d ==="
  find "$d" -type f -name '*.md' | sort
done

For each CLAUDE.md found, check whether it already has a ## Context files table:

for f in <each CLAUDE.md path from the find above>; do
  echo "=== $f ==="
  grep -A 200 '## Context files' "$f" 2>/dev/null || echo "(no context table)"
done

Keep the full list of discovered CLAUDE.md paths — Step 5 needs it to decide where each newly registered file belongs.