docs: improve coding-agent discovery

[hadrien]

Problem

• Generated agent links, canonical URLs, and discovery metadata use a placeholder documentation host.
• Agent documentation, plugin installation, and FastSQLA guidance are difficult to discover.
• Documentation pages lack compact Markdown and model-handoff controls.
• Common async SQLAlchemy failures lack searchable FastSQLA-specific corrections.

Solution

• Correct the documentation host and expose Markdown twins through metadata and compact page actions.
• Add Copy, View Markdown, ChatGPT, and Claude actions with accessible interaction targets.
• Use each Markdown twin URL in identical ChatGPT and Claude prompts.
• Package existing skills for Claude Code and Codex marketplaces.
• Add Context7 rules, package keywords, and verified troubleshooting guidance.
• Document agent resources after License and omit controls when Markdown twins do not exist.

fs-1rc

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.36%. Comparing base (4d94e0c) to head (d9630bc).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #107   +/-   ##
=======================================
  Coverage   97.36%   97.36%           
=======================================
  Files           1        1           
  Lines         114      114           
  Branches        2        2           
=======================================
  Hits          111      111           
  Misses          3        3           

Flag	Coverage Δ
fastapi-post121	97.36% <ø> (ø)
fastapi-pre121	97.36% <ø> (ø)
sqlmodel	97.36% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

[hadrien]

Independent adversarial review found two actionable documentation issues in the packaged pagination skill:

• the basic example omitted fastsqla.lifespan, leaving the injected session unbound
• min_page_size was described as a validation minimum, although runtime validation remains ge=1

Commit 686a8d7 fixes both and documents the existing-schema/async-driver requirement. The corrected example was exercised end to end against seeded SQLite; plugin/skill validators, strict docs build, Ruff, and pytest pass locally.

[hadrien]

Updated the Markdown controls from the review feedback:

• compact 44px neutral action strip directly below the page title
• Copy for LLM and View Markdown labeled controls
• accessible icon links for ChatGPT and Claude
• both model links use: Read {markdown page URL} and answer questions about the content.
• responsive wrapping keeps the model icons together on mobile

Verified with WebKit desktop/mobile screenshots and layout assertions, Chromium clipboard interaction, strict MkDocs build, Ruff, and pytest.

[hadrien]

Follow-up from the visual review:

• compact 32px painted controls while retaining 44px interaction targets
• ChatGPT and Claude hover/focus colors with subtle brand tints
• all four actions remain on one row at a 375px viewport
• exact Read {markdown page URL} and answer questions about the content. prompt for both model links
• moved For coding agents and LLMs after License in the README

Verified with the strict MkDocs build, pytest, Ruff, the UI detector, and Chromium layout/hover/clipboard assertions.