sync-wikis

name	sync-wikis
description	Sync local customizations between scaffold and sibling wiki instances — either direction. For upstream updates from GitHub, use the upgrade skill instead.

Sync Wikis (local)

Two directions of sync exist in wikibones:

Direction	Skill	When to use
Upstream → local	upgrade	Pull latest wikibones from GitHub into this wiki
Local ↔ siblings	sync-wikis (this skill)	Sync customizations between scaffold and sibling wikis — either direction

Use this skill when you've improved the scaffold or a sibling wiki — tweaked a skill, updated shared CLAUDE.md sections, fixed AGENTS.md wording — and want those changes to propagate without touching each wiki manually. Works both ways: scaffold → siblings, and sibling → scaffold.

Scope

• Canonical home: the scaffold repo (wikibones)
• Peers: sibling directories in the same parent folder that look like wiki instances (CLAUDE.md, AGENTS.md, raw/, wiki/, .claude/skills/)
• Direction: run from the scaffold to push out, or run from a sibling to pull a good change back into the scaffold first, then push from there.

What not to overwrite

1. CLAUDE.md scope body text — domain-specific, always preserve.
2. CLAUDE.md domain concept lists under Acronyms & Concepts.
3. Target-only custom skills (e.g. import-readwise) — keep them.
4. raw/ and wiki/ content — never touch.
5. Existing MANIFEST.md statuses in target wikis.

Step 1 — discover targets

SCAFFOLD_DIR="$PWD"
PARENT_DIR="$(dirname "$SCAFFOLD_DIR")"

for d in "$PARENT_DIR"/*; do
  [ -d "$d" ] || continue
  [ "$d" = "$SCAFFOLD_DIR" ] && continue
  if [ -f "$d/CLAUDE.md" ] && [ -f "$d/AGENTS.md" ] && [ -d "$d/raw" ] && [ -d "$d/wiki" ] && [ -d "$d/.claude/skills" ]; then
    echo "TARGET: $d"
  fi
done

Read target CLAUDE.md, AGENTS.md, and skill files before editing.

Step 2 — sync skills and AGENTS.md

Safe to overwrite from scaffold

1. AGENTS.md structure and wording.
2. .claude/skills/ingest/SKILL.md
3. .claude/skills/digest/SKILL.md
4. .claude/skills/lint/SKILL.md
5. .claude/skills/ingest-tweets/SKILL.md
6. .claude/skills/sync-wikis/SKILL.md
7. .claude/skills/upgrade/SKILL.md
8. llm-wiki.md
9. serve.py
10. requirements-viewer.txt
11. templates/

Required adjustments per target

1. AGENTS.md skill catalog must reflect the skills that actually exist in that target — don't add entries for skills the target doesn't have.
2. Keep target-only skills in both filesystem and AGENTS catalog.
3. Do not overwrite target-customized files under templates/ without checking the diff first.

Step 3 — CLAUDE.md shared sections

Ensure target CLAUDE.md has these scaffold-standard sections:

1. ## Scope — preserve target-specific body text.
2. ## Acronyms & Concepts — shared opening text; preserve any target concept list below it.
3. ## Images — with Obsidian image-resolution block.
4. ## Manifest — section present.
5. ## Workflows — ingest paragraph ending with: **Mark the file as \ingested:` in MANIFEST.md.**`

Merge in shared sections; do not replace the full file.

Step 4 — MANIFEST.md

Each target wiki should have a root-level MANIFEST.md. If missing, create it and list existing raw/ files with blank status:

# MANIFEST

Ingest tracker for `raw/`. Statuses: *(blank)* = not yet ingested | `ingested:<slug>` | `skip` | `merge:<slug>`

| File | Status |
|------|--------|

Step 5 — verify

for target in <target1> <target2>; do
  echo "=== $target ==="
  grep -q "^## Scope" "$target/CLAUDE.md" && echo "  Scope ✓" || echo "  Scope ✗"
  grep -q "^## Acronyms" "$target/CLAUDE.md" && echo "  Acronyms ✓" || echo "  Acronyms ✗"
  grep -q "^## Manifest" "$target/CLAUDE.md" && echo "  Manifest ✓" || echo "  Manifest ✗"
  grep -q "Resolving Obsidian" "$target/CLAUDE.md" && echo "  Obsidian images ✓" || echo "  Obsidian images ✗"
  grep -q "ingested:<slug>" "$target/CLAUDE.md" && echo "  Ingest workflow ✓" || echo "  Ingest workflow ✗"
  [ -f "$target/MANIFEST.md" ] && echo "  MANIFEST.md ✓" || echo "  MANIFEST.md ✗"
done