# Loveworld Songs

> Helps the user learn the Loveworld Singers worship songs from loveworldlyrics.com and loveworldmusic.org. Use this whenever the user mentions learning, practising, memorising or reviewing Loveworld songs, asks what to learn next, asks for a song's section order or structure, wants to find a song in the catalogue, asks about duplicate or badly formatted song pages, or wants to track progress through the 1,481-song catalogue. Also use it for "what's due for review" or "mark this song as learned", even if the user doesn't say "skill" or "catalogue". It keeps the song index and the learning progress only; it never stores or reproduces lyrics.

- Skill: `elvancedonzy/loveworld-songs` (Agent Skill, multi-file: 10 files)
- Install (CLI): `npx get-skill add elvancedonzy/loveworld-songs`
- Raw SKILL.md: https://api.skillmd.com/api/skills/elvancedonzy/loveworld-songs/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: elvancedonzy (https://skillmd.com/u/elvancedonzy)
- Updated: 2026-11-17
- Page: https://skillmd.com/skills/elvancedonzy/loveworld-songs

---


# Loveworld songs: learning helper

This skill helps the user work through the Loveworld song catalogue. It holds the
index (slugs, links, status, section structure) and the user's learning
progress, and runs a spaced-review schedule. It does not hold lyrics.

## The lyrics rule

Lyrics are copyrighted and must not be copied into this skill, into `data/`, or
into chat. Don't write lyric text into any file or reply, and don't quote more
than a fragment. When the user needs the words, send them to the song's page
(`https://loveworldlyrics.com/song/<slug>`) or to the extension's panel for
ProPresenter. Section structure ("Verse / Chorus / Bridge", slide counts) is
index data and is fine to share.

The skill folder must hold no lyric files. `scripts/install.py` refuses to copy
the skill if it finds any `.txt` file inside it.

## Setting up

The skill runs from `~/.claude/skills/loveworld-songs` (on Windows,
`%USERPROFILE%\.claude\skills\loveworld-songs`). To set it up, run
`python scripts/install.py` from the skill folder. It copies the skill there and
keeps existing learning progress. With `--link` it links to the folder instead,
for a folder synced between computers. Python 3.8 or newer is the only requirement,
on Windows, macOS or Linux (use `python3` where `python` is not found).

Run the scripts as `python "$HOME/.claude/skills/loveworld-songs/scripts/songs.py" <command>`.
Keep the double quotes and `$HOME`: that form works in bash, zsh and PowerShell,
while `~` is not expanded for Python in Windows PowerShell. Use `python3` where
`python` is not found.

## Files

- `data/catalogue.csv`: 1,481 rows copied from
  `loveworld-lyrics-extension/reports/song-check-2026-10-31.csv`. Columns: `#` (row
  number), `slug`, `url`, `status`, `section labels`, `slides`, `one-line slides`,
  `groups`, `note`. No lyrics.
- `data/progress.json`: the user's learning progress, created on first use. Once the
  skill is installed, every copy of `songs.py` uses the installed skill's file
  (`stats` prints its path), so progress made through Claude and from a terminal stays in one place.
- `data/loveworldmusic-only.csv`: 18 songs listed on loveworldmusic.org but not on
  loveworldlyrics.com, with their loveworldmusic.org links.
- `data/song-links.csv`: every distinct song link, 1,499 in all (1,481 on
  loveworldlyrics.com, 18 on loveworldmusic.org), with title and status. The
  loveworldlyrics.com titles are rebuilt from the slug, so they're approximate.
  The `loveworldmusic.org` column gives the same song's loveworldmusic.org link
  when there is exactly one match (973 songs), "N versions" when either site has
  several versions of the title (454), and is blank when the song isn't there
  (54). Matched by title from loveworldmusic.org's A–Z list on 1 Nov 2026.
- `data/added-songs.csv`: songs the user added with `add`, created on first use.
  Like progress, it lives with the installed skill and is the user's own file.
- `scripts/open_song.py`: type part of a song name, choose the match, and it
  opens that page in the user's browser, asking which site when the song has a
  loveworldmusic.org match too. It doesn't fetch or print any lyrics.
- `data/duplicate-review.csv`: generated by `dupes`. One row per song in a
  duplicate cluster, with a `cluster` id. `dupes` and reinstalling overwrite it,
  so don't record decisions here.
- `scripts/install.py`: copies the skill into `~/.claude/skills/loveworld-songs`.
- `scripts/songs.py`: the helper. Standard library only. Run it with
  `python "$HOME/.claude/skills/loveworld-songs/scripts/songs.py" <command>`.

## Commands

| Command | What it does |
|---|---|
| `stats` | Counts: songs, with lyrics, no lyrics yet, progress. |
| `check` | Validates the catalogue, and compares it with the expected numbers from `SUMMARY.md` (hardcoded in the script for 31 Oct 2026). Exits 1 on any problem. Run it after any change to `catalogue.csv`. |
| `search TEXT` | Finds songs whose slug or rebuilt title contains TEXT. Spaces, apostrophes and dashes are ignored, so `he's here` finds `he-s-here`. |
| `show SLUG` | Prints one song's index row and progress. |
| `next [N]` | The next N songs with lyrics that haven't been started. Default 10. |
| `mark SLUG learning\|learned\|skip` | Sets progress. `learned` schedules the first review for tomorrow. |
| `review SLUG good\|again` | Records a review. Successive "good" answers schedule the next review 3, 7, 14, 30, then 60 days out. "again" resets it to 1 day. |
| `due [YYYY-MM-DD]` | Learned songs due for review on that date (default today). |
| `dupes` | Rebuilds `data/duplicate-review.csv` (about 4 seconds). |
| `audit` | Writes `data/structure-review.csv`: songs whose section labels may be wrong (no chorus, all Verse, Misc, mostly one-line slides, very long, 1–2 slides), with the reasons. These are warning signs from the 31 Oct 2026 snapshot, not confirmed mistakes. |
| `extra` | Lists the 18 loveworldmusic.org-only songs and their links. |
| `add LINK ["TITLE"]` | Tracks a new song from its page link, so it shows in `search`, `next`, `mark`, `show` and `open_song.py`. Fetches nothing. loveworldmusic.org links need the title in quotes. Refuses a link that's already tracked, and a title that's already tracked unless `--anyway` is given (a different version), so the same song isn't learned twice across the two sites. |

When the user mentions a new song that `search` can't find, offer `add` with the
song's page link. Learn each song from one site: loveworldlyrics.com for the
catalogue, loveworldmusic.org only for songs the catalogue doesn't have.

To read a song, run `python "$HOME/.claude/skills/loveworld-songs/scripts/open_song.py"`,
type part of the name, and pick the match. The page opens in the browser. This is
one page at a time, read on the site, and the skill never fetches or stores lyrics.

## Learning routine

1. Start with `next 5` and open each song's URL to read and sing through it.
   Use `mark SLUG learning` when starting, and `mark SLUG learned` when the
   user can sing it from memory.
2. Each day run `due`. For each song the user sings well, run `review SLUG good`.
   If they struggle, run `review SLUG again`.
3. For a section-order question ("what comes after the chorus in X?"), read the
   `groups` column. For the words, point to the song page.

Keep sessions short. Reviewing five songs a day beats one long session.

## Duplicates and formatting

The extension already cleans up the site's messy pages: it drops repeated
copies, removes singer and repeat cues, and caps verses at 8. This skill checks
the output at the index level. It cannot judge the words themselves.

- `dupes` finds 171 clusters that share a title (for example `amazing-god`,
  `amazing-god-2`, `amazing-god-3`; a suffixed slug only counts when its bare
  title is also in the catalogue) and 96 near-identical pairs (for example
  `everday` / `everyday`, `saviour` / `savior`). 267 clusters in total.
- A cluster where every song has the same slide count and groups is a likely
  duplicate. `same_shape_as_cluster = yes` marks these; the script reports 198.
  Check one pair by opening both URLs before deciding.
- Clusters where the shapes differ are probably different versions or
  different songs. Keep both unless the user says otherwise.
- The 12 "God of Heaven and Earth" pages each have 260 slides. That's a
  very long anthem, but the same title appears on 12 pages, so ask the user
  which page to learn from.
- 32 catalogue rows have a note. 27 say "two slightly different transcriptions":
  the site has two versions and the extension keeps one. Check the page before
  learning these. The other 5 say "singer/label cue words removed".
  SUMMARY.md gives the same split.
- `check` catches structural problems. It can't catch a wrong word or a line
  in the wrong place. When the user reports one, open the song page and
  compare it with the panel.

Don't delete or merge catalogue rows. When the user decides a page is a
duplicate they won't learn, record it with `mark SLUG skip`.

## Known gaps

- loveworldmusic.org's lyrics library lists 1,501 entries, which is about 1,224
  distinct titles. Checked on 31 Oct 2026 by title:
  - 1,206 of those titles match a catalogue slug exactly.
  - 18 titles are on loveworldmusic.org only. Each returns 404 on
    loveworldlyrics.com. They're listed in `data/loveworldmusic-only.csv`, with links
    (`extra` command). Those lyrics come from loveworldmusic.org, not this catalogue.
  - 46 catalogue slugs are not listed on loveworldmusic.org. Most are the
    "hs-oct-2026" Healing Streams series, plus `king-of-eternity-pn25`,
    `medley-7-praise-night-18` and the October communion compilation.
  - The catalogue has 1,481 rows, but the `-2`/`-3` rows are repeats of the same
    titles, so the count of distinct songs is smaller.
  - The loveworldmusic.org list is loaded in the browser, so this comparison is
    a snapshot. Re-check if the library changes.
- Titles are approximate. The catalogue has slugs only, so titles are rebuilt
  from them (for example `he-s-here` shows as "He's Here"). The site has the
  real titles.
- The slide counts are a snapshot from 31 Oct 2026. If the site changes, the
  numbers go stale and the catalogue needs a fresh check.

