birdhouse-agentic-tests

Runs browser-driven end-to-end agentic tests against Birdhouse sandbox environments.

36|5|Updated Feb 27, 2026
One-click install
npx skills add https://github.com/Birdhouse-Labs/birdhouse --skill birdhouse-agentic-tests-birdhouse-labs
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: birdhouse-agentic-tests
Source: https://github.com/Birdhouse-Labs/birdhouse/tree/main/.agents/skills/internal/birdhouse-development/birdhouse-agentic-tests
Command: npx skills add https://github.com/Birdhouse-Labs/birdhouse --skill birdhouse-agentic-tests-birdhouse-labs

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Validating Birdhouse end-to-end requires exercising the real UI in a real browser against a live sandbox, which is tedious and error-prone to do manually. This Skill standardizes how agentic tests are found, executed, recorded, and scored so results are consistent and reproducible. ## Core Features & Use Cases - Browser-driven test execution: Runs markdown-defined tests against sandbox1 or sandbox2 through a real browser session, never through direct agent tool calls. - Suite orchestration: Delegates each test to a child agent, collects structured verdicts, and writes a consolidated report.md with pass/fail details. - Artifact capture: Records video, captures screenshots, and validates recording integrity before declaring a run successful. - Use Case: Ask to run the full agentic test suite against sandbox1; the orchestrator spawns child agents per test, gathers verdicts, and produces a suite report with videos and failure details. ## Quick Start Run all agentic tests in the tests directory against sandbox1 and generate a suite report.

Frequently Asked Questions about birdhouse-agentic-tests

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run Birdhouse agentic tests?▼

List the markdown test files in the tests directory, pick the target environment (sandbox1 by default), perform the sandbox ownership preflight, then follow the test's browser-driven steps. Each run produces a verdict, video recording, screenshots, and a run directory of artifacts.

How do I add a new agentic test to the suite?▼

Copy tests/template.md to a new file, fill in every section with browser-achievable steps and UI-verifiable pass criteria, then run it standalone once to confirm it is followable before committing. Tests must consume runner-provided values rather than hardcoding ports or paths.

Why can't agentic tests call Birdhouse agent tools directly?▼

The test runner lives in the production Birdhouse environment, separate from the sandbox under test, so direct agent tool calls would hit production instead of the sandbox. Every test must interact with the sandbox exclusively through the browser UI.

What is the difference between sandbox1, sandbox2, and isolated test runs?▼

sandbox1 (port 50200) is the default persistent environment, sandbox2 (port 50220) is used only when explicitly requested, and isolated runs create a fresh disposable instance via the isolated-birdhouse-web-testing skill. The choice depends on whether you need a shared stateful environment or a clean one.

Why does a test run fail even when the steps completed?▼

A run fails if the video recording is missing, zero bytes, or structurally invalid, since recording integrity is verified before declaring success. Runs also fail if any pass criterion in the test file is not met or a fail criterion is triggered.