Fix writing instructions not obeyed by LLM (v0.5.3)

Writing instructions were framed as a weak "Additional instructions:"
block at the end of the system prompt. Models prioritized competing
format directives in each stage's user prompt and ignored the style
guide. Now: system prompt wraps instructions as "MANDATORY STYLE GUIDE"
with override language, every stage user prompt reinforces compliance,
polish stage explicitly preserves lists and links, and chief editor
revision preserves structural formatting.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

CHANGELOG.md

@@ -5,6 +5,24 @@ All notable changes to PRAutoBlogger will be documented in this file.
 Format follows [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project uses [Semantic Versioning](https://semver.org/).
 
+## [0.5.3] — 2026-04-20
+
+### Fixed
+- **Writing instructions not obeyed by LLM.** The user-configured writing
+  instructions (bullet points, hyperlinks, style rules) were appended to the
+  system prompt as a weak "Additional instructions:" afterthought. Models
+  prioritized competing format directives in the user prompts and ignored
+  the style guide.
+  - System prompt now frames writing instructions as a "MANDATORY STYLE GUIDE"
+    with explicit override language.
+  - Every stage's user prompt (outline, draft, single-pass) now includes a
+    reinforcement line: "Follow EVERY requirement from your system prompt
+    style guide."
+  - Polish stage explicitly preserves bullet points, numbered lists, and
+    hyperlinks instead of flattening them.
+  - Chief editor's revision rules now preserve structural formatting (lists,
+    links) from the original draft.
+
 ## [0.5.2] — 2026-04-20
 
 ### Added