Run any Skill in Manus with one click

nclawzero-testing

Stars0

Forks0

UpdatedMay 4, 2026 at 16:55

Testing guide for nclawzero. Covers unit tests with Vitest (1700+ tests), the shard-based test harness for recording and querying results, E2E tests for ZeroClaw sandboxes, deploy scripts for test targets, and the Pi test target inventory. Use when running tests, diagnosing failures, recording test runs, or deploying to test targets.

Installation

Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.

Run Skill in Manus

Source

ncz-os

ncz-os/meta

View GitHub Repository View Creator Repositories

Download

Run Skill in Manus

Related occupationsSOC

Based on SOC occupation classification

Software Quality Assurance Analysts and TestersComputer and Mathematical Occupations·SOC 15-1253

SKILL.md

readonly

Run any Skill with one click

name	nclawzero-testing
description	Testing guide for nclawzero. Covers unit tests with Vitest (1700+ tests), the shard-based test harness for recording and querying results, E2E tests for ZeroClaw sandboxes, deploy scripts for test targets, and the Pi test target inventory. Use when running tests, diagnosing failures, recording test runs, or deploying to test targets.

nclawzero Testing Guide

How to run, record, and analyze tests for nclawzero. Covers unit tests, the test harness, E2E tests, and test target deployment.

Unit Tests

nclawzero has 1700+ unit and integration tests organized into three Vitest projects (defined in vitest.config.ts):

Project	Scope	Path
`cli`	CLI integration tests	`test/*/.test.{js,ts}`
`plugin`	Plugin unit tests (co-located)	`nemoclaw/src/*/.test.ts`
`e2e-brev`	Cloud E2E (requires `BREV_API_TOKEN`)	`test/e2e/brev-e2e.test.js`

Running All Tests

npx vitest run

Running a Specific Project

# CLI tests only
npx vitest run --project cli

# Plugin tests only
npx vitest run --project plugin

Running a Single Test File

npx vitest run test/credentials.test.ts

Watch Mode

npx vitest

Test Harness

The test harness (scripts/test-harness.py) is a shard-based test result storage and query CLI, modeled after the MNEMOS memory management system. It records Vitest results as JSON shards and supports search, comparison, and trend analysis.

Storage location: ~/.nclawzero-harness/runs/

Recording a Test Run

python3 scripts/test-harness.py record

This runs the full test suite, captures the output, and stores results as a timestamped JSON shard.

Record a specific suite:

python3 scripts/test-harness.py record --suite cli
python3 scripts/test-harness.py record --suite plugin

Querying Results

Command	Description
`python3 scripts/test-harness.py stats`	Summary across all recorded runs
`python3 scripts/test-harness.py recent 10`	Most recent 10 runs
`python3 scripts/test-harness.py failures`	Failures from the latest run
`python3 scripts/test-harness.py failures RUN_ID`	Failures from a specific run
`python3 scripts/test-harness.py flaky`	Detect flaky tests (intermittent pass/fail)
`python3 scripts/test-harness.py flaky 14`	Flaky tests over a 14-day window
`python3 scripts/test-harness.py compare RUN1 RUN2`	Diff two runs side by side
`python3 scripts/test-harness.py trend`	Pass rate trend over time
`python3 scripts/test-harness.py trend 30`	Pass rate trend over 30 days
`python3 scripts/test-harness.py search "credential"`	Search test names and error messages
`python3 scripts/test-harness.py get RUN_ID`	Full details of a specific run

Exporting and Maintenance

# Export all runs
python3 scripts/test-harness.py export json
python3 scripts/test-harness.py export csv
python3 scripts/test-harness.py export markdown

# Prune old runs
python3 scripts/test-harness.py prune 90    # Delete runs older than 90 days

E2E Tests

ZeroClaw E2E

The primary E2E test proves the complete ZeroClaw user journey: install, onboard with the zeroclaw agent, verify the sandbox, and run live inference.

NEMOCLAW_NON_INTERACTIVE=1 \
NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE=1 \
NVIDIA_API_KEY=nvapi-... \
  bash test/e2e/test-zeroclaw-e2e.sh

Prerequisites:

Docker running
NVIDIA_API_KEY set (real key, starts with nvapi-)
Network access to integrate.api.nvidia.com

ZeroClaw Sandbox Operations

Tests gateway operations against a running ZeroClaw sandbox:

bash test/e2e/test-zeroclaw-sandbox-operations.sh

Test cases:

TC-ZSO-01 — Health probe (/health)
TC-ZSO-02 — API status (/api/status)
TC-ZSO-03 — Webhook inference (/webhook)
TC-ZSO-04 — Config reload after restart
TC-ZSO-05 — Gateway restart and recovery
TC-ZSO-06 — Dashboard accessible (GET /)

Environment variables:

Variable	Default	Description
`ZEROCLAW_HOST`	`localhost`	Gateway host
`ZEROCLAW_PORT`	`42617`	Gateway port
`NEMOCLAW_SANDBOX_NAME`	`e2e-zeroclaw`	Sandbox name

Stub-Based Testing (Offline)

For CDN-blocked networks (like NVIDIA OmniStation), use stub images:

# Build stub images (one-time)
scripts/build-stub-images.sh

# Run with stubs
nemoclaw onboard --agent zeroclaw

# Verify (stub marker present)
curl http://localhost:42617/health
# {"status":"ok","version":"0.6.9-stub","stub":true}

Deploy Script

scripts/deploy-test-target.sh deploys and runs the ZeroClaw E2E suite on any test target. It installs Docker and Node.js if missing, clones or updates the repo, runs the full test suite, and copies results back.

Targets

# SSH to a remote host
scripts/deploy-test-target.sh --ssh pi@10.0.0.56 --nvidia-key nvapi-xxx

# Brev cloud instance
scripts/deploy-test-target.sh --brev awesome-gpu-name --nvidia-key nvapi-xxx

# Local machine with live images
scripts/deploy-test-target.sh --local --nvidia-key nvapi-xxx

# Local machine with stub images (no network required)
scripts/deploy-test-target.sh --local --stub

Options

Flag	Description
`--ssh USER@HOST`	Deploy via SSH (key auth)
`--brev INSTANCE`	Deploy via Brev
`--local`	Run on this machine
`--token TOKEN`	GitHub PAT for private repo access
`--nvidia-key KEY`	NVIDIA API key for live inference
`--branch BRANCH`	Branch to test (default: `nclawzero-rebase`)
`--stub`	Use stub images (no Docker Hub required)
`--skip-install`	Skip Docker/Node install
`--keep`	Keep container after test
`-v, --verbose`	Print all remote output

Test Targets

nclawzero maintains Raspberry Pi test targets for edge validation:

Target	IP	Role	Description
zeropi	10.0.0.56	Minimal footprint	ARM64 Pi for baseline resource testing
clawpi	10.0.0.54	Full-featured	ARM64 Pi for full ZeroClaw feature testing

Deploying to a Test Target

# Deploy to zeropi (minimal footprint)
scripts/deploy-test-target.sh --ssh pi@10.0.0.56 --nvidia-key nvapi-xxx

# Deploy to clawpi (full-featured)
scripts/deploy-test-target.sh --ssh pi@10.0.0.54 --nvidia-key nvapi-xxx

Results are written to ./e2e-results/ on the invoking machine (configurable with --results-dir).

Repository Topology for Testing

ARGONAS bare repo                    — LAN source of truth
  /mnt/datapool/git/nclawzero.git
     |
  zeropi (.56) ~/nclawzero           — min footprint test target
  clawpi (.54) ~/nclawzero           — full-featured test target

The deploy script clones from ARGONAS when deploying to Pi targets.

Related Skills

nclawzero-zeroclaw-get-started — Installation and first run
nclawzero-zeroclaw-config — Configuration reference
nclawzero-zeroclaw-security — Security hardening reference
nclawzero-skills-guide — Skills catalog for nclawzero

name	nclawzero-testing
description	Testing guide for nclawzero. Covers unit tests with Vitest (1700+ tests), the shard-based test harness for recording and querying results, E2E tests for ZeroClaw sandboxes, deploy scripts for test targets, and the Pi test target inventory. Use when running tests, diagnosing failures, recording test runs, or deploying to test targets.