v0.1.13

Fixed

• The WP Super Cache adapter no longer poisons the cache. It is now a
  total bypass, not a separate cache bucket. On a GET/HEAD request carrying
  Accept: text/markdown it hooks WP Super Cache's cache_init action, sets
  $cache_enabled = false and defines DONOTCACHEPAGE. Phase 1 returns before
  serving and wp_cache_postload() returns before the output-buffer callback is
  installed, so the request neither reads nor writes the page cache.
  The previous marker-based design (appending a fixed marker in
  wp_cache_get_cookies_values) gave markdown requests their own key but also
  made them eligible to be written under it; a marked request that fell
  through to HTML rendering — no cached .md yet, activation before backfill
  completes, new content — was cached as HTML under the markdown key and then
  served to every later markdown request carrying the same marker and encoding.
  This produced blocked sites and a gzip-specific blind spot (the key includes
  the encoding, and wp_remote_get requests gzip by default). The
  wp_cache_get_cookies_values marker hook is removed entirely; no marker, no
  separate key, no encoding dependence. Registration by path via
  wpsc_add_plugin(), gating on the markdown toggle, deactivate/uninstall
  removal, and the Expert (mod_rewrite) caveat are unchanged.
  • Marker-keyed entries left on disk by earlier versions become unreachable to
    markdown requests (the read path is skipped) and expire on their own; no
    cleanup code is added.
• Self-test probe selection and encodings. Query-string permalinks are
  excluded from the probe set (a query string bypasses the cache being tested,
  so such a URL could return markdown on a site whose clean paths are blocked;
  if fewer clean candidates exist, fewer are probed — a query URL is never
  substituted). Each chosen URL is now probed in both Accept-Encoding
  variants (gzip and identity), because the cache key differs by encoding;
  working requires every reachable probe in both encodings to return markdown,
  and any HTML is blocked with the URL and encoding named in the detail.