📋 Changelog

All 5 posts had meta descriptions cut off mid-sentence (20–70 chars).

These are thin WordPress taxonomy listing pages (58 tag pages, 31

The _redirects cross-domain rule was missing the protocol prefix on

Refactor add_reading_time_indicator to iterate over all

elements and inject a per-article reading-time + word-count span into the article's .entry-meta. Skips list/archive cards without .entry-content, avoids duplicate insertion, strips non-content tags, requires a minimum content length, and computes minutes based on ~200 wpm. Also added a couple of Bash patterns (grep for entry-content in scripts and python3) to .claude/settings.local.json.

- Add fef9726a-c61d-45a1-a624-61e63214bc70.txt to public/ so

Tighten exception handling and add safer defaults across scripts. Replaced several bare excepts with specific exception tuples to avoid silencing unexpected errors (markdown_exporter.py, wp_spell_check_and_fix.py, wp_to_static_generator.py). Added defensive JSON parsing with fallbacks/breaks for invalid responses in WP fetch logic. Switched image cache hashing from MD5 to BLAKE2b (digest_size=16) for stronger, fixed-size hashes (optimize_images.py). Hardened the search overlay by adding a CSS class selector, sanitizing item URLs (fallback to '#') and updating close/overlay selectors to target the new class (search.js). Also narrowed datetime parsing exceptions to specific types. These changes improve reliability, error visibility, and security of the tooling.

Update workflow and scripts to make caching and cache-detection more reliable, and fix Cloudflare purge endpoint.

Several improvements across CI, Cloudflare Worker template, and asset tooling:

Add JSON-LD validation in content_validator: warn on missing structured data for article/main pages, detect empty or invalid JSON-LD blocks, ensure @context references schema.org, and require headline/datePublished/author for Article-like types. Refine CSS preload heuristic in enhance_html_performance: only preload stylesheets whose filename matches critical/main/styles (.min optional) to avoid preloading unrelated styles. Improve meta description filler in fix_seo_issues: scope first-paragraph extraction to article/main or common content containers to avoid grabbing nav/footer/sidebar text.

Workflow: bump image build-cache key to v3, include .build-cache.json in cache, and add human-friendly cache status output. Add a cleanup step to remove ./static-output after failed builds to avoid stale artifacts.

CI: capture the spell checker exit code into SPELL_EXIT, derive a human-friendly SPELL_STATUS, and append that status to the GitHub step summary (retain non-blocking behavior). Worker: make Cache-Control dynamic by using getTTL(path) instead of a fixed max-age=300. HTML optimizer: refine optimize_images to treat the first image inside

or

as the LCP candidate (fetchpriority="high"), set loading="lazy" for other images, add decoding="async" for all images, update the docstring, and remove the prior index-based heuristic.

Add Gzip fallback to the static site compression pipeline and improve HTML/CSS validations and parsing.

All 164 existing article/page HTML files had relative paths (e.g.