Add agentic product testing: agents verify the product as real users

[0xhackerfren]

Summary

• Adds a new verification layer (Phase 5b) where AI agents operate the assembled application as real users with defined profiles and goal-oriented objectives
• Captures UX quality metrics (click count, navigation depth, backtracking, error encounters, confusion index) with quantitative UX budgets as quality gates
• This is a uniquely agentic capability: no previous testing methodology could instruct a test to 'figure out how to accomplish this goal' and measure whether the experience was efficient and discoverable

New files

• reference/product-testing.md: Full reference covering user profiles, test objectives, UX budgets, experience signals, pipeline integration, agent behavior, metrics, and feedback to specification
• harness/templates/product-test-template.md: Template for defining product test scenarios with objectives, budgets, and spec traceability
• harness/templates/user-profile-template.md: Template for defining simulated user personas with technical literacy, domain knowledge, constraints, and behavioral tendencies

Updated files

• MANIFESTO.md: Product testing added to Section VI (The Machine) -- the assembled product verified by agents acting as users
• harness/PLAYBOOK.md: Phase 5b (Product Testing) between Review and Deployment, with inputs, outputs, exit criteria, human/agent responsibilities, quick reference row
• harness/templates/test-plan-template.md: Product test section for defining objectives alongside module-level molds
• reference/workflow.md: Product testing as step 7 in the pipeline, flow diagram updated, kanban stations table updated, companion list updated
• reference/glossary.md: Four new terms (product test, test objective, user profile, UX budget)
• harness/.cursor/rules/codex-automata.mdc: Product testing section and terminology
• README.md: Product testing section
• CHANGELOG.md: Unreleased entries
• ROADMAP.md: Current section

Test plan

• Verify reference/product-testing.md cross-references resolve
• Verify templates are self-contained (GitHub URLs for references outside harness)
• Verify glossary terms are in alphabetical order
• Verify workflow.md flow diagram includes product testing
• Run markdownlint CI