blog-spell-check-loop

name	blog-spell-check-loop
description	Spell-check the wickstrom.tech blog by iteratively running Bombadil against the local live-reload server, triaging flagged words, and fixing the source posts, dictionary, or extractor until clean.
user_invocable	true

Prerequisites

• The blog must be running locally at http://localhost:8080 (live-reloading dev server).
• spellcheck.ts, wordExtractor.ts, and custom.utf-8.add live under wickstrom.tech/ (a subdirectory of the bombadil-playground CWD). All commands below use paths relative to the project root — do not cd.
• Markdown posts to edit live in ~/projects/wickstrom.tech/src/posts/ (a separate checkout — not under the playground).
• URL pattern: posts are served at the root, e.g. http://localhost:8080/2020-07-02-the-todomvc-showdown-testing-with-webcheck.html — not under /posts/, even though the markdown sources live in src/posts/. The filename (minus .md) becomes the slug.

Loop

Repeat until step 1 finds no flagged words:

1. Run the spell check

mkdir -p .tmp
bombadil test --headless --exit-on-violation --time-limit 5m \
  http://localhost:8080 wickstrom.tech/spellcheck.ts \
  > .tmp/blog-spellcheck.txt 2>&1

Exit code 2 = violation found. Exit code 0 / timeout = clean run on the pages Bombadil explored in 5 minutes (which may not be exhaustive — see step 3).

Bombadil prints violations directly on stdout — read .tmp/blog-spellcheck.txt to collect the flagged words and the URLs where they were found.

2. Triage each flagged word

Three outcomes. Apply the right one per word, not per run.

True positive (real typo). Find the post under ~/projects/wickstrom.tech/src/posts/ and fix the markdown source. Do not add typos to custom.utf-8.add.

False positive — legitimate word the dictionary doesn't know. Two options, choose by context:

• Add to custom.utf-8.add when the word is reusable across posts (proper nouns, recurring jargon, brand names, regional spellings). One word per line, no count header.
• Mark inline with spellcheck="false" when the word is local to one post (a one-off code-like token, a stylized name, a quoted handle). Prefer adding the attribute to an existing inline element (<code>, <span>, <a>); only wrap in a new <span spellcheck="false">…</span> if no suitable element exists. Block-level wrapping (<div>) is a last resort.

False positive — extraction noise. The token shouldn't have been spell-checked at all (URL fragment, decoded entity remnant, hash, etc.). Fix wordExtractor.ts rather than polluting the dictionary. After any changes there:

npm --prefix wickstrom.tech test

Add a unit test covering the new filter case before fixing.

3. Verify the fix on the specific failing page

A 5-minute exploration doesn't guarantee Bombadil revisited the page you just edited. After applying fixes, target each previously-failing URL directly with a short time limit:

bombadil test --headless --exit-on-violation --time-limit 10s \
  <failing-url> wickstrom.tech/spellcheck.ts

Use 5s for a quick smoke check, 10s if the page is large. If this still flags the same word, the fix didn't land — re-triage. If it's clean, move on.

4. Loop

Go back to step 1. The full 5-minute run may surface words on pages the previous run never reached.

You're done when step 1 completes its 5-minute window with no violations.

Reporting

At the end, summarize:

• True positives fixed (word -> correction, with file path).
• Words added to custom.utf-8.add.
• Inline spellcheck="false" additions (file + element).
• Any wordExtractor.ts changes (one-line rationale each).

Keep it terse. The user reads the diff.