[SeanShaffer88]

Reorganizes the repository so an AI assistant can use it as a reference without ingesting fake data or the whole 2.3 MB catalog.

• Archives 1,000 generated placeholder entries (index/master-manifest.json, the seven categories/01–07 manifest.json files, manifests/resources.json) and the incompatible registry schema to archive/legacy-template/, with a README stating they are not real resources and must never be cited or merged. The old AI_INDEX.md (which described a modules/ layout that never existed) is archived there too.
• Rewrites AI_INDEX.md as an accurate one-page index: 12 domains, counts, and which file to open for each.
• Removes the template header block from AGENTS.md and gives it a smallest-first reading order.
• Adds derived AI views generated by the new scripts/generate_ai_views.py: index/resources.compact.json (full catalog, shortlisting fields only, ~38% of the canonical size), 97 per-category shards in index/by-category/, and index/category-groups.json rolling the 97 categories into 12 domains. resources.json remains canonical and is unchanged.
• README gains a TL;DR pointing AI readers to AI_INDEX.md; llms.txt, index/README.md, and CONTRIBUTING.md updated to match (including the new regeneration step).

Validation: python scripts/validate.py passes (2,202 records, 97 categories), unit tests 5/5, and no placeholder URLs remain outside archive/.

[SeanShaffer88]

Enhancement packs A–C are folded into this PR (4 additional commits):

Pack A — freshness. New scripts/link_check.py (reachability checker, stdlib only) and a monthly link-check workflow (added via the web UI, as workflow files require). Backfill of the 306 records that had no recorded link status: 293 confirmed live and recorded with checked_on 2026-10-28, 0 dead, 2 moved Android documentation URLs updated everywhere they appear, and 11 left untouched because they failed only from the build machine's network (checker-limited, e.g. Page Flows, Behance, Awwwards).

Pack B — alternatives graph. New scripts/generate_related.py derives related-entry links from shared topics/use cases within each category: records with related entries go from 24 to 1,335, links symmetric as the validator requires. Empty topics were filled on 37 records under a strict rule (a topic must be carried by at least 60% of the category); the rest were left blank rather than guessed.

Pack C — thin entries. 100 short purpose lines expanded with a fit/constraint clause, propagated to the category and catalog pages. The remaining 119 obscure entries were left as-is rather than padded with guesses. The 4 cross-listed URLs (Catch2, google/benchmark, Tracy, WAMR) were verified as already grouped by project_id — project-groups.json covers non-GitHub endpoints only, so no change was needed. Adds llms-full.txt (one-file edition, generated by scripts/generate_ai_views.py).

Validation after all packs: python scripts/validate.py passes (2,202 records, 97 categories) and unit tests pass 5/5. The repository homepage has also been set to the Dev Resource Explorer.