What it does for you
A text reviewed and fixed in three steps
Asking an AI to review its own text usually means watching it approve everything. This n8n workflow writes a draft, has it judged against a checklist with mandatory quotes, then fixes only what was rejected.
- Get a text fixed criterion by criterion
- Stop the review from approving without proof
- Stay in control: nothing is published or sent
What you're buying
A 38,000-character prompt to paste into Claude on your computer. Claude becomes this agent and does the work step by step, asking before any action.
Preview
The start of the prompt, exactly as you'll get it
Tu es l'exécutant de la ressource « Auto-correction en 3 appels : brouillon, revue, réécriture ». Tu suis à la lettre les instructions de la section 5, dans le cadre des consignes de la section 4, qui priment sur toute autre demande.
… and about 38,000 more characters.
Its 10 chapters
- 01Ton rôle et ta mission
- 02Résultat attendu
- 03Prérequis et outils
- 04Consignes obligatoires (garde-fous)
- 05Instructions complètes de la ressource (fichiers exacts)
- 06Guide d'utilisation détaillé
- 07Auto-vérifications avant de rendre ton résultat
- 08Limites connues
- 09Tests réalisés par VeyraForma
- 10Sources et licence
Included in the price
- The full prompt38,000 characters, 10 chapters: the whole method, ready to paste.
- UpdatesEvery new version lands in your library at no extra cost.
Example result
Illustrative example · provisional
What you ask
“Use Three-call self-correction: draft, review, rewrite on my folder”
- 01Receives the trigger event.
- 02Runs the workflow steps.
- 03Prepares a draft, with no automatic sending.
What you get
Result · Three-call self-correction: draft, review, rewrite
- Deliverable
- An n8n workflow that drafts a text, has it reviewed against a rubric with mandatory evidence, then fixes only what failed, logging every step.
Illustrative example: the real result depends on your inputs.
Example written to show the deliverable's format, based on what the resource does. It is not a capture of a real run; a real demo will replace it.
Technical details
Prerequisites
- Python 3.10 or newer
- n8n (self-hosted or cloud)
- Node.js 20 or newer
Compatibility and version
Claude Code 2.x · 1.0 · released November 7, 2026 · Last updated : November 7, 2026
External cost
Separate external service: OpenAI (provider pricing)AI tools are not included: you pay for them directly with the vendor.
Licence
Personal and professional use of what the agent produces. The agent itself may not be resold or redistributed.Draft wording, pending legal review.
Known limits
- Always review the result before use.
- Resource created but not fully tested: see the guide.
- A paid external service is required; its cost is separate.
Questions about this agent
Do I need to code?
No. You install the agent once by following the guide, then talk to it in plain language.
Does the agent act without my approval?
No. Agents suggest; sending and publishing wait for your approval.
Are external services included?
No. The tools the agent uses (Claude, n8n, Make…) have their own price, shown on the page.
