2 / 2 · Book 1 · What Are Agent Skills? · drill: interview Q&A← prev⊞ allGet the book →
1.12Interview Depth Check
Question 1
Prompt: Explain the difference between "intent" and "implementation" in browser test automation. Why does this distinction matter? A strong answer covers:
- Intent is the purpose ("user can log in and reach the dashboard"); implementation is the mechanism (
click('#submit-btn'),waitFor('.dashboard')) - When a selector changes, intent is unchanged but implementation breaks; traditional tools see only implementation, so the test fails
- Agents bridge the gap: they take a fresh snapshot, find the button labeled "Submit," and click it by ref
- UI details change constantly; business logic rarely does. Intent-driven tests break only when the logic breaks -- which is when you want them to
Question 2
Prompt: Describe the six-layer architecture of the skill-based stack. What happens if a layer is removed? A strong answer covers:
- Test Definitions: without them, no repeatable, versioned specifications
- AI Agent with skill loaded: without it, no reasoning or adaptation; without the skill, the agent does not know the CLI exists
- Bash Tool: without it, the agent can reason but cannot execute -- intelligence without action
- playwright-cli: without it, the agent writes raw Playwright code, paying tokens for boilerplate and losing disk-first snapshots
- Playwright engine and browser: without them, no reliable waiting/actionability, and nothing to automate
Question 3
Prompt: Why has SKILL.md become a de facto cross-agent standard, and what does that mean for QA? A strong answer covers:
- Near-zero barrier: a skill is a markdown file with YAML frontmatter -- no compilation, server, or API key
- Portability: one file works in Claude Code, Cursor, Gemini CLI, and Codex CLI, so teams are not locked to a vendor
- Vendor adoption closed the loop:
playwright-cli install --skillsmeans the tool ships its own teaching material; skills.sh adds discoverability - The generated skill is commodity infrastructure; the differentiator is architecting frameworks around it -- test design, self-healing, CI, token optimization