infra(seo): ship llms.txt + llms-full.txt, upgrade RSS feed, refresh site description

[julianken]

Context

Source: docs/seo-strategy/phase-4/analysis-report.md, Gate 3 — Amplify (discovery infrastructure).

Ship three discovery-surface improvements:

1. llms.txt + llms-full.txt — major AI engines mostly ignore (~10% adoption), but Cursor / Continue / Cline IDE agents actively consume these. That's the audience.
2. RSS feed metadata upgrade — <author>, <copyright>, <generator>, <managingEditor>, <image> at channel level; per-item <author>. Improves Inoreader/Feedly visibility and AI scraper ingestion.
3. Site description rewrite — current "A tech blog and reference catalog on agentic AI." is generic; replace with a more specific, citation-friendly description.

Acceptance criteria

• New file public/llms.txt (static, ~30 lines) — minimal disclaimer + indexed-content pointers per analysis report §3.7
• New file src/app/llms-full.txt/route.ts — Next.js route handler returning Markdown index generated from PATTERNS (filter !archived) + getPublishedPosts(). Reuse getPublishedPosts from src/lib/queries/posts.ts. Content-Type: text/plain; charset=utf-8. Cache-Control: s-maxage=3600, stale-while-revalidate=86400
• src/app/feed.xml/route.ts — add channel-level fields: <author>{CONTACT_EMAIL}</author>, <copyright>, <generator>Next.js 16 / Payload CMS</generator>, <managingEditor>, <image> (using og-default.png). Add per-item <author> using CONTACT_EMAIL. Defer <content:encoded> — requires a Payload Lexical → HTML server-side exporter that doesn't exist (audit finding); document the deferral in PR body as a follow-up issue
• src/lib/site-config.ts — update siteDescription from "A tech blog and reference catalog on agentic AI." to: "A diagnostic analysis of agentic AI design patterns in practice — 24 reference patterns, field reports from production agentic workflows, and the gap between what agents promise and what they deliver."
• src/app/(frontend)/layout.tsx — verify the new siteDescription flows through metadata.description (no edit needed if already imported)
• Defensive CONTACT_EMAIL add: if src/lib/site-config.ts doesn't already export it (B might have added it first), add it here. Idempotent. Trivial rebase if both PRs land
• Local verification (after pnpm dev):
  • curl http://localhost:3000/llms.txt returns 200 + plaintext
  • curl http://localhost:3000/llms-full.txt returns 200 + Markdown listing 24 patterns + 4 posts
  • curl http://localhost:3000/feed.xml | xmllint --noout - validates as well-formed XML
• pnpm lint && pnpm test:unit && pnpm typecheck pass
• PR body includes Closes #<this-issue> + curl output of the three endpoints + note about <content:encoded> deferral
• julianken-bot review posted with APPROVE verdict

Out of scope

• <content:encoded> (full-body RSS) — requires Lexical → HTML server exporter; separate follow-up
• Other gate 3 items (schema fields, posts, npm package) — separate issues
• Tier C identity — separate issue

Reference files

• Plan: ~/.claude/plans/make-a-plan-to-peppy-swan.md — Category E
• Analysis report: docs/seo-strategy/phase-4/analysis-report.md — Gate 3.7 (llms.txt), Gate 3.8 (RSS)
• Technical infra audit: docs/seo-strategy/phase-1/area-2-technical-infra.md — RSS gaps + llms.txt rationale

Workflow

• Branch: infra/seo-discovery-surfaces
• Worktree: ../wt-{this-issue}-seo-discovery-surfaces
• Conventional commit: infra(seo): llms.txt + RSS metadata upgrades + site description
• PR title: same as commit
• Subagent type: general-purpose

[julianken]

Starting work on #390 (Wave 1 / Category E). Worktree: /Users/j/repos/wt-390-seo-discovery-surfaces. Branch: infra/seo-discovery-surfaces.

[julianken]

PR opened: #394

Pre-PR verification:

• pnpm lint ✅
• pnpm test:unit ✅ (525/525)
• pnpm typecheck ✅
• curl /llms.txt ✅ (200, plaintext)
• curl /llms-full.txt ✅ (200, lists 24 patterns + 4 posts)
• curl /feed.xml | xmllint --noout - ✅ valid XML

Deferred: <content:encoded> for RSS pending Lexical exporter.

Awaiting julianken-bot review.

[julianken]

Loop applied on PR #394:

• IMPORTANT (channel-level ): removed; covers the role per RSS 2.0 spec.

The bot's SUGGESTION about siteDescription duplicates in schema config + manifest is plan-controlled (issue scoped only site-config.ts) — leaving for a follow-up issue.

Verification:

• Feed XML well-formed (xmllint)
• pnpm lint ✅
• pnpm test:unit ✅ (525/525)
• pnpm typecheck ✅

Awaiting julianken-bot re-review.

[julianken]

Loop 2 applied on PR #394:

• IMPORTANT (RSS oversize): downscaled android-chrome-192x192.png to 144x144 via sharp; new asset at public/rss-icon-144.png; block updated with explicit /.
• Regeneration script committed at scripts/generate-rss-icon.ts.
• siteDescription duplicates in schema config + manifest remain plan-controlled (issue scoped only site-config.ts).

Verification:

• file public/rss-icon-144.png → 144x144 PNG ✓
• pnpm lint / pnpm test:unit / pnpm typecheck ✅
• Feed XML still well-formed

Awaiting julianken-bot re-review.