[Task] SEO/GEO: tidy GitHub discoverability assets (topics, social preview, README support)

[Astro-Han]

SEO/GEO follow-up item 4 from PR #1189. Review feedback recommended splitting this into an independent PR.

Value assessment

Value: low to medium, with low implementation cost and several quick wins. GitHub is the only acquisition channel currently verified as effective from the LLM Wiki analysis, so making repository assets send a consistent discoverability and entity signal is worth the small effort.

Note: the repository social preview image cannot be set through the API. It must be uploaded manually in repository Settings, or called out as a manual follow-up.

Cost: low.

Goal

The repository description, topics, social preview, and README structure send one consistent entity signal, and README / website / llms.txt wording does not drift.

Scope

In scope:

• Keep topics at 20 or fewer, prioritizing real search terms such as desktop-ai, ai-agent, open-source, macos, windows, chatgpt-oauth, local-ai, document-ai, office, ollama, lm-studio, and opencode; compare against the current 19 topics and trim where needed.
• Add small Support / Community / Maintainers sections to README so it answers where to get help and who maintains the project.
• Prepare the repository social preview image at 1280x640, reusing og-image if appropriate. Upload remains a manual Settings step.
• Align the README cover image alt text with the entity description.
• Align the release notes template so macOS signed / Windows unsigned status is consistent across README, website, and llms.txt.

Out of scope:

• Website changes covered by other issues.

Relevant files or context

• README.md / README_CN.md
• site/public/og-image.png, if reused for the social preview
• Repository Settings for topics and social preview; some of this may require manual action

Verification

• Topics are 20 or fewer and all are real search terms.
• The top 30 lines of README answer what PawWork is, why it is useful, how to start, where to get help, and who maintains it.
• Platform and signing language matches llms.txt and the website.

Execution mode

Investigate and propose a plan first - post the topic-trimming proposal and README change list, then wait for an explicit approved comment before editing. Social preview upload must be handled manually.