Browser Performance Validation

---
name: browser-performance-validation
description: Build repeatable AgentWorks browser performance validation with customer-defined budgets and comparable samples. Use for page, journey, Web Vital, resource, or timing regressions.
---

# Browser Performance Validation

## Outcome

Create repeatable browser performance measurements for approved pages and journeys, compare them with customer budgets or baselines, and retain enough provenance and diagnostics to explain regressions.

## When to use

Use after browser setup for performance budgets, trend tracking, or release checks. Treat results as environment-specific observations unless the customer provides a controlled benchmark environment.

## Discovery and user direction

Inspect the current workflow, goals, metrics, configuration, capabilities, stores, reports, and triggers before proposing changes. Summarize reusable design and gaps, then ask focused questions for unresolved customer choices such as scope, success, approvals, thresholds, ownership, and budgets. Record the answers as customer direction. Installation alone does not approve workflow changes or execution. Default to one small-team workflow; split only for incompatible access or lifecycle boundaries.

## Required inputs

Resolve target pages/journeys, tested build, browser/device/network profile, cold/warm-cache policy, sample count, metrics, budgets or baseline, acceptable variance, evidence retention, and pass/fail policy.

## Plan and AgentWorks tools

Use scripted steps for controlled execution, metric collection, aggregation, comparison, and persistence. Use a message sequence to investigate supported regressions. Keep warmup and measured samples explicit; do not let an agent invent thresholds after seeing results.

## Knowledge and persistence

Store raw samples, aggregates, budgets, environment/profile identity, and comparison results in durable DB tables. Store traces, video, console/network logs, and exported profiles in durable assets. Put reusable application performance facts in KB notes only after verification.

## Validation and reporting

Require the configured number of valid samples and comparable provenance. The dashboard shows the customer's chosen aggregates, variance and budget status, filters for journey/browser/device/build/environment, and drill-down to samples, traces, evidence, missing metrics, invalid samples, and history. A partial measurement set cannot silently pass.

## Guardrails

Do not claim production performance from an uncontrolled test environment, mix incompatible device/network/cache profiles, change budgets to obtain a pass, or ignore functional failures. Redact sensitive URLs and network data before retention.

## Read details when needed

- [Workflow design and outcomes](../../references/workflow-design-and-outcomes.md): goals, metrics, and current-versus-separate workflow decisions.
- [Performance measurement](../references/performance-measurement.md): shared provenance, sampling, budgets, and comparison rules.
- [Browser QA plan and tools](../../browser-qa/references/agentworks-plan-and-tools.md): browser step and platform-tool choices.
- [Evidence capture](../../browser-qa/references/evidence-capture.md): durable browser diagnostic artifacts.
- [Performance guide](references/performance-workflow.md): sampling, comparison, storage, and acceptance cases.
- [Example budget](examples/performance-budget.json): fictional policy shape.
- [Catalog metadata](playbook.json): presentation and optional recommendations.

## Completion contract

Return installed playbook and budget revisions, customer overrides, tested environment/build/profile, sample completeness, metric decisions, actual trial result, evidence/report locations, capability resolution, and limitations.