GuideGenerating autoreg.md

Generating autoreg.md

autoreg.md is the spec that tells AutoReg about your application, and it’s generated for you — not written by hand. The generate-autoreg AI skill reads your app and produces a complete autoreg.md with confirmed selectors and realistic test data, ready to import.

Tests generated from a skill-written spec typically pass on the first run, because every selector in the spec was confirmed to exist in your app before the file was written.

Step 1 — Install the skill

Install the AutoReg: Skill Installer extension (autoreg.autoreg-skill-installer) in VS Code (or any VS Code-compatible editor). It installs in seconds into any repository, application, or project:

  1. Open the AutoReg panel in the activity bar.
  2. The extension detects which AI coding agents exist in your workspace — Claude Code, Cursor, GitHub Copilot, Windsurf, Continue, Cline, Aider, Gemini CLI, Codex CLI, and more.
  3. Click Install next to your agent.

Using the skill:

  • Agent detected — type the slash command in your coding agent’s chat: 
    /generate-autoreg
  • No agent detected — a generate-autoreg/ folder is created in your workspace with the skill inside it. Reference it directly in your agent’s chat: 
    @generate-autoreg/generate-autoreg.md

Step 2 — Run the skill in your agent

In Claude Code, for example:

/generate-autoreg /path/to/autoreg-workflow /path/to/app/src
  • First argument — your AutoReg workflow folder (where autoreg.md will be written). Optional if you run it from inside the workflow.
  • Second argument — the root of the application to analyze.

Step 3 — Answer the interview

Before it gets started, the skill asks you up front for things it can’t work out on its own:

  • Test credentials — accounts for each role
  • Seed data — records that already exist in the test environment
  • Sample values — realistic inputs for forms (names, amounts, dates)

Secrets you provide are kept out of the spec’s plain text and resolved securely at import time.

Step 4 — Review the generated spec

The skill writes an autoreg.md containing:

SectionWhat’s in it
Confirmed selectorsOnly selectors that were confirmed to exist in your app — no guesswork.
Every routeEvery page and API route in your app, including which ones require login.
Realistic test dataValues that match your app’s actual forms and validation rules, so tests exercise real scenarios and real error messages.
Page behavioursLoading states, modals, toasts, and other UI patterns the generator should expect.
Auth flowsLogin, logout, registration, and the roles your app supports.
Test Data CatalogThe credentials, seed data, and sample values used across the suite.
8–25 test scenariosPositive and negative cases, ready to generate tests from.

AutoReg automatically verifies every spec the skill produces when you import it, so you can trust it’s genuine and unaltered.

Step 5 — Import into AutoReg

Create a new AutoReg workflow (or open an existing one) pointing at the folder containing the generated autoreg.md. On import, AutoReg:

  1. Verifies the spec — confirms it came from the skill and hasn’t been changed since.
  2. Resolves any pending secret values — you’re prompted for them, and they’re stored securely rather than written into the spec.
  3. Starts test generation — see Generating tests.

Refreshing a spec

If your app changes, re-run the skill rather than editing autoreg.md directly — that keeps the spec accurate and lets AutoReg continue to recognize it as genuine.

Troubleshooting

ProblemFix
Skill not offered in the agentRe-open the AutoReg panel and click Install; check the autoreg.autoDetect setting is enabled.
Spec doesn’t import as verifiedRe-run the skill to produce a fresh copy rather than editing the file by hand.
Selectors in the spec don’t match the deployed appThe deployed version differs from what was analyzed. Re-run the skill against the branch that is actually deployed.

Next steps

What is autoreg.mdGenerating Tests