gasyoun/agents-skills-public

cologne-pages-landing

Use this skill when the user asks to run the migrated source command cologne-pages-landing.

Command Template

/cologne-pages-landing <REPO|all>

Created: 06-07-2026 · Last updated: 05-09-2026

Default path (seconds — run this first)

python ~/Documents/GitHub/Uprava/tools/cologne_pages_landing.py --dry-run

LLM / judgment residue only after the script report exists (or the script exits non-zero with a real defect).

Give every Cologne dictionary repo a proper GitHub Pages front door: a single self-contained index.html (SEO meta + Open Graph + JSON-LD Book schema + styled hero/CTA layout) plus .nojekyll so Pages serves it statically.

Default path — run the script, don't hand-roll discovery again. Uprava/tools/cologne_pages_landing.py (H2790, Sonnet 5 — ZJ pages-landing deploy) is the mechanical port of Phases 0/3 below: it set-diffs local sanskrit-lexicon/* clones for a missing SEO landing, applies the same csl-*/tooling skip-rules, and — for any repo whose facts are already filled into its DICT_FACTS table — opens the PR via the same chassis as cologne_batch_deploy.py.

cd ~/Documents/GitHub/Uprava
python tools/cologne_pages_landing.py --dry-run          # set-diff, read-only (default)
python tools/cologne_pages_landing.py --selftest         # fixture + GRA/MD template round-trip
python tools/cologne_pages_landing.py --pr --repo <CODE> # open the PR (needs DICT_FACTS[<CODE>] filled first)

A repo the script reports NEEDS_FACTS has no index.html but no DICT_FACTS entry yet — do Phase 2 below by hand (or LLM), add the entry to DICT_FACTS in the script, then re-run --pr --repo <CODE>. A repo reported NEEDS_REVIEW has a real non-SEO index.html — read it before touching anything; the script never overwrites it.

Gold-standard reference implementations (structure only — do NOT redesign):

• ~/Documents/GitHub\GRA\index.html (German-title dict)
• ~/Documents/GitHub\MD\index.html (English-title dict)

The phases below are the manual/fallback procedure the script encodes — read them to understand why the script does what it does, or when the script itself needs a fix, not as the everyday runbook.

---

Phase 0 — Target discovery (script does this — --dry-run)

Argument = one repo code (e.g. SKD) or all.

Exclusions (mirrored in the script's TOOLING_SKIP set + csl-* prefix rule):

• Skip tooling/infra repos (csl-pywork, csl-orig, csl-corrections, csl-apidev, csl-websanlexicon, cologne-data, csl-* generally) — this skill is for dictionary repos whose Pages site is a landing page, not an app.
• Skip repos with a non-SEO index.html that is a real app/site (check before overwriting — if index.html exists but has no og:site_name, READ it first; only replace trivial redirect/placeholder pages, never a functioning site).
• csl-orig is never touched (org rule).

Phase 1 — Publish-safety gate (cheap form)

These repos are already public and the landing page contains only already-public bibliographic facts, so the full /publish-safety-check is normally not needed. Run it ONLY if the target repo is private or the page would surface anything beyond title/author/year/description/scan-link.

Phase 2 — Gather per-dictionary facts

For each target, collect (from the repo README, prefaces/, and CDSL):

Fact	Example (GRA)
Full title	Wörterbuch zum Rig-Veda
Code badge	GRA
Author + place, year	Hermann Grassmann · Leipzig, 1873
One-sentence description	"…complete concordance-dictionary of the Ṛgvedic vocabulary…"
Canonical URL	https://sanskrit-lexicon.github.io/GRA/
Browse URL	https://www.sanskrit-lexicon.uni-koeln.de/scans/GRAScan/2020/web/webtc/indexcaller.php
License	CC-BY-SA-4.0 (per RH1 policy; check the repo's actual LICENSE)

Verify the Browse URL actually resolves (HTTP 200) before baking it in — the <CODE>Scan/2020/web/webtc/indexcaller.php pattern holds for most but not all dictionaries. If Cologne is unreachable, check Uprava/SERVER_OUTAGES.md before concluding the URL is wrong.

Phase 3 — Deploy (script does this — --pr)

Per target repo:

1. Write index.html from the gold-standard template with the facts substituted (title/desc in <title>, meta description, OG, Twitter, JSON-LD, hero). Keep og:image = https://sanskrit-lexicon.github.io/cdsl-card.png. The script's TEMPLATE/build_landing_html() does this from DICT_FACTS.
2. Write empty .nojekyll at root (this was a follow-up fix commit in every first-wave repo — bake it in from the start).
3. Opens a PR on branch cologne-pages-landing/<CODE> via the cologne_batch_deploy.open_file_pr chassis (GitHub-API writes, not a local commit+push — see that module's docstring for why).
4. Check Pages is enabled: gh api repos/sanskrit-lexicon/<REPO>/pages. If 404, enable from the default branch root:

   gh api repos/sanskrit-lexicon/<REPO>/pages -X POST \
     -f "source[branch]=<default>" -f "source[path]=/"
   

5. After Pages builds, verify the live URL returns the new page (poll a couple of minutes max; if the build lags, log it and move on).

Cap parallelism at 4. Continue past per-repo errors; log DONE/ERROR per repo.

Phase 4 — Roll-up + hygiene

• Emit a table: | Repo | Result | Live URL | Commit |.
• Append to each touched repo's .ai_state.md ## ✅ Completed: - DD-MM-YYYY: /cologne-pages-landing — SEO landing + .nojekyll live.
• If this was an all run, note remaining skips (real-app index.html repos, unreachable scan URLs) so the next run knows they were deliberate.

---

Notes for future maintenance

• The page is one self-contained file: inline CSS, no JS, no external assets except the shared OG card image. Keep it that way.
• If the shared design changes, update GRA/MD first (they are the reference), then re-run this skill in explicit-replace mode (a human decides when).
• /cologne-onboard-repo should include this skill for any newly adopted dictionary repo.

Dr. Mārcis Gasūns