Maurice, for your code?
Your coding assistant reads your repository and tells you where Maurice would help, which first pilot to run, and where it brings nothing. Read only. Nothing to install.
Copy this text into your coding assistant, opened in your project folder. It gives its diagnostic in a few minutes, in the conversation.
Read https://agentmaurice.ai/skill/maurice-diagnostic/SKILL.md and follow it to run a read-only Maurice diagnostic of this repository.From which assistant?
| Assistant | How |
|---|---|
| Claude Code | Open Claude Code in the project folder and paste the text. |
| Codex | Open Codex in the project folder and paste the text. |
| Cursor and others | Paste the text into the agent's chat, with the project open. |
What the assistant looks for
- AI model calls repeated on a rule-based task → classify, extract, check: Compiled AI runs them as code, for almost nothing.
- Business scripts, cron jobs and batches → scheduled Workflows.
- Home-made agents → agent mode, plus compiled Workflows for the rule-based steps.
- Bots and webhooks → channels: messaging, HTTPS API, MQTT.
- Small internal interfaces and hand-coded validations → MiniApps and human approvals.
- OCR, RAG, scraping, SSH scripts → the AgentMaurice MCP servers, open source on GitHub.
- Directory, rights, shared files rebuilt by hand → the Company in a Box Modules.
- SaaS already connected → connected services, with a Console account.
What you get
- The verdict: high, medium or low potential.
- Ranked opportunities, each with the file and line that justify it.
- One small first pilot, with a measurable success criterion.
- Where Maurice brings you nothing, said plainly.
- A JSON block the assistant reuses to build the pilot once Maurice is installed.
What it does not do
- It changes no file, except the report if you ask for it.
- It opens neither your .env files, nor your keys, nor your secrets.
- It sends nothing to AgentMaurice. Your code stays between you and the assistant you already use.
- It invents no figures: it only cites our published measurement, with its limits.
The Skill
The diagnostic is a public Skill, published on GitHub. Your assistant can read it online, or you can copy it into the Skills folder of Claude Code or Codex. Direct link:
What next?
Like the pilot? Install Maurice from the same assistant, or let's talk.
Are you an assistant? Read the page for agents.
Frequently asked questions
- Do I need to install Maurice?
- No. The diagnostic needs no binary, no account and no key.
- Is my code sent anywhere?
- Nothing is sent to AgentMaurice. The assistant you already use reads your repository, under its own rules.
- Is it free?
- Yes. Only your assistant usage counts, as for any other task.