Quickstart¶

Get your first adversarial test report in under five minutes.

Step 1 — Point teststop at your project¶

Navigate to any software project:

cd /path/to/your/project

Or stay in your home directory and pass the path:

teststop run --path /path/to/your/project

teststop works on any project in any language. No setup needed.

Step 2 — Run¶

teststop run

teststop will:

Scan your project (language, type, routes, dependencies)
Load any existing confidence memory
Compose the adversarial mandate
Call your AI CLI (claude or copilot)
Display the test scenarios and confidence report

The first run takes 20–60 seconds depending on your AI CLI.

Step 3 — Read the output¶

A typical run looks like:

teststop v0.1.0 — adversarial user testing

Project:   my-api (Go · api)
Adapter:   claude
Depth:     normal
Scenarios: 9

SCENARIO  TS-001 · critical
  Title:   Double-submit on slow network
  User:    Someone with a poor connection clicking Submit twice
  Steps:
    1. Fill checkout form
    2. Click submit (simulate 3s latency)
    3. Click submit again before response
  Failure modes: Duplicate order created, inventory decremented twice

SCENARIO  TS-002 · high
  Title:   Session cookie after password change
  ...

──────────────────────────────────────────────────────────
Confidence:  0%  (0 areas tested — first run)
Areas:       volatile: auth, checkout, api
Exit code:   1 (REVIEW)
──────────────────────────────────────────────────────────

Exit 1 on the first run is normal

The first run always exits 1 (review required) because confidence starts at 0%. As you run teststop repeatedly and scenarios keep passing, confidence builds and eventually exits 0.

Step 4 — Run again¶

teststop run

teststop remembers what it tested. Confidence grows with each clean run. After ~15 passes per area, that area is considered proven and retired.

Check your accumulated confidence state:

teststop status

Area         Confidence   Tests   Maturity   Status
──────────────────────────────────────────────────
auth         62%          5       growing    active
checkout     38%          2       new        active
api          19%          1       new        active

Step 5 — Commit your memory¶

git add .teststop/memory.json
git commit -m "chore: add teststop confidence memory"

.teststop/memory.json is the accumulated proof that your system works. Commit it so your team shares the same confidence baseline.

Common flags¶

# Machine-readable output for CI / AI agents
teststop run --output json --no-color --quiet

# More thorough testing
teststop run --depth aggressive

# Custom confidence threshold (default 80%)
teststop run --threshold 90

# Show the exact mandate sent to the AI
teststop mandate --show

# See full command reference
teststop --help

Next Steps¶

Understand the pipeline

Learn how teststop scans, composes, and generates scenarios end to end.

How It Works
Memory System

How confidence scores work, how areas retire, and what to commit.

Memory System
Agent Integration

Wire teststop into Claude Code, Copilot, or any AI coding workflow.

Agent Integration
CLI Reference

Every command, flag, and option.

CLI Reference