docs: add quick start and prompting guide, move style guide into authoring

[sebastian-iancu]

Summary

Adds two human-readable docs for people who use the openehr-specs plugin, and removes docs/spec-style-guide.md.

• docs/quick-start.md is a tutorial. It clones BASE and AA_GLOBAL, installs the plugin, reviews a chapter, fact-checks class names, adds an amendment record entry and builds the HTML preview, all on a throwaway branch that the last step discards.
• docs/prompting-guide.md is a how-to organised by situation: new spec, chapters, prose, amendment record, class changes, review, preview, releases, ITS-REST, scaffold and commit messages. Each has a copyable prompt and notes on what to check.
• They are two files because a tutorial and a prompt reference are different document kinds.
• Both are linked from the root README, docs/install.md and the plugin README.

Style guide. It restated conventions that already live in content-patterns, authoring and review, and it was a second copy to keep in step. The three conventions no skill carried (admonitions, code blocks, tables) move to skills/authoring/references/asciidoc-syntax.md, which authoring and content-patterns point to. The four links to the old file are removed, and the prompting guide is added to the inventory-sync rule in AGENTS.md and CONTRIBUTING.md.

Reviewer notes

• Two one-line edits in AGENTS.md: the style-guide bullet is removed, and docs/prompting-guide.md is added to the list of docs that move with the skill inventory.
• Tone, structure and writing-principles sections of the old guide were not copied, because the skills already state them. They are in git history if you want them back.
• scripts/validate.py does not check these docs. The skill and subagent names in the prompting guide are an inventory kept by hand.

Checklist

• python3 scripts/validate.py and python3 -m unittest discover -s scripts -p 'test_*.py' pass
• Skill triggering tested locally in Claude Code (see docs/testing.md). The review and amendment-record steps were run through headless Claude, and the Docker build and cleanup steps on a temp clone. The identifier-grounding step, marketplace install and a full run from a clean machine were not.
• Cursor install tested locally when plugin or manifest content changed. Not applicable to the manifests (unchanged), and the tutorial's slash commands were not tried in Cursor.
• Both marketplaces updated when plugin metadata changed (no metadata changed)
• CHANGELOG.md updated under Unreleased; no version bump
• Content grounded in published openEHR specifications, no invented identifiers. Behavioural claims about each skill were checked against its source with the prose-reviewer agent, and its seven major findings are fixed.

🤖 Generated with Claude Code