Stage 00 — User task

Actualizado a septiembre 2026
Con licencia
esDisponible en ES
Pagos rápidos
Solo 18+
Índice de contenidos
  1. Input
  2. Why this stage exists
  3. Actions
  4. Output
  5. Gate criterion
  6. Report

Input

Why this stage exists

Search intent across different iGaming queries differs radically. One keyword leads to a TOP-list of operators, another to a map of payment options, a third to a regulatory breakdown, a fourth to a "how to do X" instruction. There is no universal template. If you start straight from pool cleanup and the plan, the article ends up formally "covering the clusters" but failing to solve the task the user came to search with.

This stage is first and decisive. It articulates what exactly the user wants to receive on opening the page. All subsequent stages (plan, writing, validation) work toward this articulation, not abstract "informational/commercial intent."

Actions

1. SERP reconnaissance (mandatory)

Via web_search, get the first ten results for KEYWORD in the ЯЗЫК ВЫДАЧИ. Then via web_fetch pull 3 pages from the top — pick the most representative ones (skip Wikipedia, skip forums — real topical pages on the subject).

From this reconnaissance extract:

1a. Cross-check with geo-clusters and legality in the GEO (mandatory)

This is a safeguard against silent cluster drops. If the keyword pool or the KEYWORD itself contains:

then the model must perform an additional check: exactly which locally licensed (i.e. licensed in the relevant GEO) options for this topic actually exist in reality as of 2026. This is done via an additional web_search of the form " + license " and, when needed, web_fetch of the regulator’s registry (whitelist, license register, etc.).

For each locally licensed option found — a brief relevance assessment against the user task:

Ban on silent drops: if the pool contains a geo cluster or a legality cluster, and the final user-task formulation and verification questions contain not a single local option (even as an "imperfect" answer), this is a stage failure. The local regulatory picture must be honestly reflected in the user task: either as "no such options exist" or as "there are N such options with such-and-such limitations."

2. User task formulation

Based on the SERP reconnaissance, articulate 2–3 sentences about what task the reader is solving with this article. The formulation must be concrete, no abstractions.

Bad (too generic):

Good (concrete, verifiable):

3. Main semantic node

In one sentence: the core value that the reader should walk away with. Not a restatement of the task — the single most important thought or outcome the article exists for.

4. Verification questions (3–5)

Formulate 3–5 concrete questions the article must answer directly. These are not stage-02 FAQ — they are a test for closing the user task. They are used in final validation at stage 06.

Examples for different topics:

5. SERP justification

A short (2–3 sentences) fixation of what exactly the top analysis showed and why the formulated task is correct. This is insurance against a crooked formulation: if the agent’s intuition suggested one thing, but the actual top shows another — the actual top wins.

6. Priority for subsequent stages

Record the explicit rule: any formal constraints (volume, TOP-N operators, mandatory blocks, summary tables, FAQ distribution) apply only if they serve the solution of the user task. If a rule conflicts with the task — the task wins. If a pool cluster does not work toward the task — it is closed compactly.

Output

Create the file _workspace/00_intent.md with the structure:

Gate criterion

The stage is considered passed if:

If any item fails, the stage is re-run addressing the failed item. Limit — 3 attempts.

Report

After passing the gate, one line in the chat: Stage 00 done: user task recorded, verification questions N, local GEO options O.

Escrito por los editores de «Casinos en Vivo España».