Full Check

Full check

Goal

Do end to end what the individual skills do piecemeal: measure, fix, verify, and keep watching — with one price agreed before anything is spent. This skill orchestrates the others; it adds no tool of its own, and its output is one report: what the site scored, what changed, what it costs to know more.

Required inputs

Ask, in one message, only what is missing:

1. The domain, and the page that matters most (default: homepage).
2. What they sell and to whom, in one line — it seeds the keyword work.
3. A buyer question the user wrote, if they have one — required for the visibility legs, never invented.
4. Whether the site's source is in the working directory (fixes get applied) or not (fixes get listed).

Before spending

Quote the tiers with real prices from citable_prices, let the user pick one, and treat that yes as consent for every call listed in the tier — including any single call over $0.20 — and for nothing else:

Standard and full need the user-written buyer question; without one, run page tier and offer competitor-citations ($0.65) to surface real questions. Anything beyond the tier — more pages, more keywords, extra visibility runs — is a new ask at its listed price.

Workflow

1. Baseline. citable_domain_overview ($0.03) for size and traffic; citable_onpage_audit on the money page ($0.01). Standard and up: citable_ai_visibility on the user's question ($0.05/engine), citable_keyword_research on the seed ($0.06), citable_rank_check on the top 5 keywords by volume ($0.012 each). Save the numbers — score, positions, engines citing — to a small JSON in the repo (next to rank-tracking's history file if one exists): this is the before.
2. Fix. Work the audit's topFixes[] in order, exactly as citability-audit does. Where a fix has a build skill, use it: ai-crawler-access for robots decisions, llms-txt for the map, schema-markup for structured data, geo-rewrite when headings and answers are the problem. All free. No source in the working directory → deliver the ordered fix list instead and stop after step 3.
3. Demand read (standard and up). From the keyword research: which keywords the page should target, where the domain already ranks, and — full tier — what AI cites the rival for (competitor-citations, ~$0.65) and what that implies for this site's pages.
4. Deploy gap. Verification only means anything against the live (or preview) URL. If deploy happens in-session, continue; if not, report the baseline + applied fixes and say exactly what to run after deploy: re-audit ($0.01) and, standard and up, the same visibility check ($0.05/engine).
5. Verify. Re-run the audit and the visibility question. Report before → after: score delta, engines that cite the domain now versus before, positions if enough time has passed for them to mean anything (say so if it hasn't).
6. Leave monitoring behind. Offer the monitor skill on the keywords and question from this run — a scheduled weekly rank + AI-citation watch with a hard per-run budget. That is the standing version of everything this run measured once.

Close with the report: baseline, changes, after, total spent (sum the _payment.transaction receipts), and the one next thing worth paying for.

Guardrails

• The tier consent is the spend ceiling. Nothing outside the picked tier runs without a new price-and-ask.
• Never invent the buyer question — no question, no visibility leg.
• One ai_visibility run per leg is a snapshot, not a trend: engine answers vary between runs. Say so; 2–3 runs ($0.05/engine each) are an add-on, not a default.
• Don't re-run any call whose result this run already has; the verify legs exist because the site changed, not to double-check the feed.
• If the audit's js-content check fails, say the verify legs will keep failing until rendering is fixed — do not spend on visibility checks for a page crawlers cannot read.