agentsclimarketplace

Se ranking analyst

Skill bogdankrupin/se-ranking-super-user/skills/se-ranking-analyst

Pull, analyze, and report on SE Ranking data — keyword rankings, backlinks, AI/LLM search visibility, SERPs, and site audits — via the SE Ranking MCP tools or API. Use this skill whenever the task touches SE Ranking data or a tracked project: weekly or monthly client SEO reports, "what moved this week", rank-tracking changes, backlinks gained or lost, AI visibility and ChatGPT/Perplexity brand mentions, competitor SERP comparison, or computing period-over-period deltas. Trigger even when SE Ranking is not named explicitly but the user clearly wants analysis of rankings, backlinks, AI search visibility, or audit data for a site they track. It encodes which tool to reach for per data category, how to resolve a project and comparison dates, how to compute period-over-period deltas accurately, and the report format.From its SKILL.md

Install
npx -y skills add bogdankrupin/se-ranking-super-user --skill se-ranking-analyst

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

8.0 KB, as published. Nobody here has run it

SE Ranking Analyst

Turn raw SE Ranking data into a defensible analysis: pull the right metrics, compute period-over-period change, explain what moved and why, and propose actions. The whole point of this skill is that the tool selection, the safe-delta rules, and the report shape are settled in advance, so each run is fast and consistent instead of re-derived from scratch.

Mental model: two tool families, one API

SE Ranking exposes a large set of tools, and the safe default is to load a tool's exact parameters before calling it (e.g. via tool search) rather than guessing parameter names. They split into two families, and it's worth picking the right one up front:

  • PROJECT_* — operate on the user's own tracked projects (identified by a site_id). These reflect the real tracking setup: the keywords being monitored, the search engine and location configured, the actual check dates. Use these for client reporting and anything about "our site / our tracked keywords".
  • DATA_* — research tools that work on any domain or keyword without a project. They consume SE Ranking data-API credits. Use these for competitor research, prospecting, or pulling a metric that isn't set up as a tracked project.

When the user means "my client's monthly report", that is almost always PROJECT_*. When they mean "how does competitor X look" or "research this keyword", that is DATA_*. The full categorized tool map is in references/endpoints.md — read it before pulling data so you reach for the correct tool the first time.

The same logic applies if the user is building an n8n / raw-API pipeline rather than calling MCP tools live: the endpoint families and the safe-delta rules below are identical; only the call mechanism differs.

Step 1 — Resolve the project and the comparison window

Do this before pulling anything.

  1. Identify the exact project. If you don't have a site_id, call PROJECT_listProjects and match by domain. Confirm the project name/domain back to the user before pulling — this makes sure the whole report is built on the project the user actually meant, and it is cheap to confirm now.
  2. Establish two dates: the current period and the comparison baseline. Use PROJECT_getCheckDates (dates positions were actually checked) or PROJECT_getHistoricalDates (standard comparison points like yesterday / last month) so you compare against a date that genuinely exists, not an interpolated one.
  3. Confirm scope: which data categories are in this report (rankings, backlinks, AI visibility, audit) and which search engine / location, if the project has several.

Step 2 — Pull the data

Consult references/endpoints.md for the right tool per category. Pull the current period and the baseline period for each category in the report. Keep each category's raw result separate and labelled — do not merge pulls into one blob before analysis, or the downstream reasoning gets a tangle of unattributed numbers.

Step 3 — Compute deltas SAFELY (read this every time)

These keep each report built only from its own project's real data. Treat the following as hard rules, not preferences:

  • One project per report. Never merge data from multiple site_ids into a single client's analysis. Run each client as an isolated pass.
  • Missing data is missing — never substitute. If a metric has no record for a date (SE Ranking writes no record when nothing changed, so gaps are normal and meaningful), report it as "no data for this period". Don't fill a gap with the previous period's value or with an unrelated figure. Absence is signal, not a hole to fill.
  • No silent reuse. If you reuse a prior result for any reason (e.g. a genuinely unchanged metric), say so explicitly in the output. Do not present reused or carried-over numbers as freshly pulled.
  • Compute change as current minus baseline, per metric, with both raw and percent where it helps. For positions, lower is better — a move from 8 to 3 is an improvement of 5; state direction in words so the sign is never ambiguous.

Step 4 — Analyze

Lead with what changed, not with raw tables. For each category, surface the movements that matter and skip the noise.

Default significance thresholds (state them in the report; treat as tunable, and adjust if the user has their own):

  • Rankings: a position change of 3+ spots, or any keyword crossing the top-10 / top-3 boundary in either direction.
  • AI / LLM visibility: a change in visibility share or in the count of prompts where the brand is mentioned; new prompts where the brand appeared or dropped out. This is volatile — be explicit about the period and prefer share-of-mentions over single-prompt anecdotes.
  • Backlinks: net new vs lost referring domains (weight domains over raw link count); any lost link from a high-authority domain.
  • Audit: new critical errors vs the previous crawl; issues resolved.

For each significant movement give: what moved (metric + magnitude + direction), the most plausible why (tie to a known cause when one exists — an algo update, a new/lost link, a content change — and clearly mark it as a hypothesis when it is one), and a recommended next action. Do not invent a cause to fill the slot; "cause unclear, worth investigating" is a valid and honest line.

Step 5 — Format the report

Default structure (adapt to the user's house style if they have one):

# [Client / domain] — [period] vs [baseline period]
## TL;DR
2-4 bullets: the headline movements only.
## Rankings
What moved past threshold, with direction and likely why.
## AI / LLM visibility
Share + prompt-level changes, with the period stated.
## Backlinks
Net referring-domain change; notable gains/losses.
## Site audit (if in scope)
New critical issues; resolved issues.
## Recommended actions
Prioritised, tied to the movements above.
## Data notes
Dates compared, any "no data" gaps, anything reused or carried over.

The Data notes section is not optional — it is where the safe-delta discipline becomes visible and auditable.

Client-facing delivery

Anything client-facing (an email, a posted report) is an explicit-permission action: draft it, show the user, and wait for a clear yes before sending. Do not auto-send. The human approval gate is the last line of defence, not a step to optimise away.

Credit and token discipline

DATA_* calls cost data-API credits; live MCP analysis costs tokens — two separate budgets. Pull only the categories and date pairs the report needs, batch where a bulk endpoint exists (e.g. DATA_getKeywordsMetrics for many keywords at once), and don't re-pull a period you already have in context. Keep the Step 3 data rules in place even when optimising for credits — accuracy comes first.

Language

Match the language of the request. If a deliverable is for a specific audience and the language isn't obvious, ask which language it should be in.

Keep looking

Skills are one crate of 608,475. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.