TestNeo
TestNeo Docs Studio
TestNeo › Docs › Studio Skills & Pipelines
Studio · Web AI Assistant

Studio Skills & Release Pipelines

Use TestNeo Studio to prove a claim, run the right regression, and get a PASS / WARN / BLOCK ship signal — without pasting five prompts by hand. This guide is for QA, SDETs, and release owners using the in-app AI Assistant.

Open Studio: app.testneo.ai/web → pick a project → Skills (or type /skills in chat). Local: /web/ai-assistant?projectId=<id>

What should I open?

You want…Use
Prove a requirement live, then decide to ship Trust loop pipeline
Predict + run regression, then decide to ship PR regression pipeline
Repeat your own steps every sprint Custom pipeline (+ New)
One job only (“list tests”, “validate PR”) Skill or Discover

Trust loop

Claim → evidence → gate.
Needs context + base URL. ~10–20 min.

PR regression

Predict → run → gate.
Needs tests in the project. ~3–8 min.

Honest limit: One green “Prove claim” run is not full regression proof. The release gate uses project history, suites, silent-break, and memory — not a single NLP pass.

Before you start

  1. Open your web project in Studio.
  2. For Trust loop / Prove claim: select a unified context (bugs, Figma, or requirements).
  3. Set a base URL (Environments or project URL) so Run live works.
  4. Optional: start your self-hosted agent (preferred when online; otherwise TestNeo Cloud).
  5. Hard-refresh the browser after a Studio deploy so Skills and pipelines load.

Trust loop — claim → ship

In plain English: Don’t only ask AI for more tests. Prove one business requirement works live, use memory to pick coverage, run suites, check collateral, then get PASS / WARN / BLOCK.

Steps

  1. 1Prove the claim — generate 1–3 NLP tests from context.
  2. 2Run live — execute from the Test Artifact (video / dashboard evidence).
  3. 3Have we seen this before? — Engineering Memory hot areas + suggested tests.
  4. 4Predict + run suites — Intelligent Regression Suite, then Multi-Test Runner.
  5. 5Silent-break — collateral latest failures you didn’t re-prompt.
  6. 6Can I release? — project ship signal (PASS / WARN / BLOCK).

How to start

PR regression pipeline

In plain English: Something changed (or you’re about to ship). Predict which tests matter, run them, see if anything else went red, then ask can we release?

Steps

  1. 1Predict suite — from PR/files + memory + history (or inventory cold-start).
  2. 2Run tests — predicted pack / Test Runner.
  3. 3Silent-break — collateral check.
  4. 4Release gate — Can I release?
  5. 5Explain (optional) — if WARN / BLOCK.

How to start

Run a pipeline (hands-off)

  1. Open Skills.
  2. On the Release Pipelines board, pick Trust loop or PR regression (or a custom pipeline from the dropdown).
  3. Click Run pipeline.
  4. Stay in chat — steps advance automatically. If a step needs you (Run live / pick suites), follow the pause hint, then continue.
  5. Read the Pipeline Score Report at the end.

Guide = same path with one coach button at a time (better for demos).

Don’t paste the whole playbook markdown or all step queries in one chat message — only the last command runs. Use Run pipeline or click one step at a time.

Custom pipeline (+ Custom prompt)

Build a repeatable QA path for your project — ordered skills, not an n8n free canvas.

  1. Release Pipelines → + New.
  2. Set a title (e.g. My smoke).
  3. Add skills from the left palette. Watch the live preview strip update.
  4. Optional: add Custom prompt and type any Studio skill phrase, e.g. Did I break something else?
  5. Keep Hands-off Run pipeline checked → Save.
  6. Select it in the custom dropdown → Run pipeline.

Minimal example

1. List newest tests
2. Custom prompt → Did I break something else?
3. Release gate → Can I release?

Drag the grip handle to reorder. Expand a step to edit title, prompt, or “only if WARN/BLOCK”.

Pipeline Score Report

When hands-off finishes, chat shows one summary card so you don’t re-read every step:

Use Explain / Suggest fix plan skills after WARN or BLOCK.

Skills vs Discover

SkillsDiscover
When Guided playbooks + full catalog for this project One-tap starters in this chat
Open /skills · Skills button /discover · Starters
Feel Gallery + pipelines + setup checks Fast overlay

Same engine. Different door. Prefer Skills when you need Trust loop / PR regression / custom pipelines.

FAQ

Is this the same as ChatGPT writing Playwright?

No. Studio generates runnable NLP, executes in product (agent or Cloud), and gates with release intelligence. Export Playwright is optional for git/CI — not the evidence path.

Do I need a PR number?

Not for Trust loop or a basic PR regression smoke. Use Validate PR #n (or a pipeline variable pr_number) when you want a PR-scoped Release Brief.

Why did delete ask for confirmation?

Deleting a custom pipeline cannot be undone. Studio uses an in-app confirm dialog (not a browser popup).

Where is Engineering Memory?

Skill: Have we seen this before? or Open engineering memory. Full platform guide: Release Intelligence → Engineering Memory.

Copy-paste prompts

Reuse these in chat or as Custom prompt steps:

Start trust loop
Start quick release
Predict regression suite
Have we seen this before?
Did I break something else?
Can I release?
Show release metrics
Validate PR #101
Explain PR failures
Suggest fix plan
List newest tests
Prove this business claim

Where to try it