Monthly content — tunisie-electricite.com

You are in the cms monorepo clone. One page. Per month. Shipped as a draft PR.

Authorised 2026-08-19 (register entry TN-001) after two monthly reviews asked for permission and got no answer. The authorisation is standing: the PR review is the safeguard, so you never need to ask again — but it is for one page a month, and a second page in the same month is a new question, not a judgement call.

Before choosing: grade the last one

Read memory/journal.md for the previous content entry. If its PR merged at least 28 days ago, measure the page before writing anything new:

gsc page https://tunisie-electricite.com/<path> --site tunisie-electricite.com

Journal impressions, clicks, position and the queries it actually landed on. A page that earned nothing is the most useful thing you will learn all month — say so plainly, and say what you think went wrong. Do not quietly write another one on the same pattern.

Choose from measured demand

gsc-opportunities --site tunisie-electricite.com
gsc queries --site tunisie-electricite.com --limit 200

A topic qualifies only if the demand is already visible in our own data. Three honest sources, in order:

1. page-two — queries with real impressions stuck at position 8-25. We are already shown for them; sometimes the answer is a page that deserves to win.
2. Queries with impressions and no page that targets them. The genuine content gap: cross the query list against sites/tunisie-electricite.com/pages/.
3. The latest reports/*-business-review.md — its ranked opportunities, which were reasoned about with more context than a query list carries.

Never pick a topic because it sounds useful. If you cannot point at a number, it is not a candidate. "Tunisians probably want to know about X" is how a site fills up with pages nobody searches for.

The check that matters most

Does an existing page already serve this query? Run gsc-opportunities --only cannibals and grep the page inventory before committing to a topic. In August the site had 33 queries split across two of its own URLs, and closing two of them took a full PR. A new page that competes with an existing one is worse than no page: it costs a month of work and takes clicks off a page that was already winning.

If an existing page half-covers the topic, the answer is not a new page:

• snippet or intent problem → leave it for ctr-optimizer;
• the page needs real rework → the repo's /improve skill is the right tool, but it runs on Python scripts (keyword_audit.py, improve-gate.py) and this image has no Python (node:22-bookworm-slim; verified 2026-08-21). Open a decision entry in decisions.md rather than half-doing it by hand.

Write it with the repo's own pipeline

Use the monorepo's /article skill. It owns keyword research, SERP analysis, Blade structure, the hero image, routes and verification, and it is maintained by people who ship pages every day. Do not reimplement any of it here — this skill decides what and whether, /article decides how.

Country rules that override nothing and constrain everything:

• Hard rule 1 — never invent a fact. Every tariff, phone number, address, procedure, deadline or figure comes from an authoritative source, steg.com.tn first, cited in the PR body. A number you cannot source does not go on the page, however much the paragraph wants it.
• Never restate a STEG tariff figure. Read departments/tunisia-energy/data/tariffs.php. In August the same wrong redevance lived in two pages and took two fixes; that file exists so it can never happen again. If your page needs a figure that file does not carry, add it there.
• The billing model is not intuitive — the tranche reached sets one price for the whole month's consumption, VAT is 7/13% on energy but 19% on the redevance, and the 300 kWh threshold is on the monthly average. The file's header explains it. Getting this wrong is the single most likely factual error on this site.
• French (fr-TN), matching the tone of existing pages.
• Scope: sites/tunisie-electricite.com/ and departments/tunisia-energy/ only.

Ship it

1. Branch countryos/tn/<YYYY-MM-DD>-content-<slug> from origin/develop.
2. Ship one page. If research turns up three good topics, write one and put the other two in the journal for next month.
3. Push and open a draft PR with gh pr create --draft, titled [tn] <page title>. The body states: the query the page targets and its measured impressions, why no existing page serves it (with the cannibalisation check you ran), and every source for every figure.
4. Journal the URL, the target query, its baseline impressions, and the date this becomes measurable (merge + 28 days).
5. Report the PR URL and the target query in your final message.

Never merge or mark the PR ready yourself.

Standing candidates

Carried from the July and August reviews, each still to be re-verified against current data before you commit to it:

• STEG smart meters — 500,000 planned by end 2026, the end of estimated bills, real monthly billing. No page covers it; little French-language competition. Complements /steg/facture.
• A bill calculator — demand is proven and mostly in Arabic ("how much do N kWh cost"). Already specified in proposals/calculateur-facture/SPEC.md; it is a tool, not an article, so it is a separate build.
• The outage editorial /steg/coupures — the /coupures/* pages capture the "where and when" and do it well; the "why, and what to do" page has not moved since July while summer load-shedding recurred.