TASK GUIDE · SOURCE-AWARE

Agent Skills for UI and accessibility review

A practical review stack for responsive UI, accessibility checks, and visual evidence.

Build this plan for my Agent
Desired outcome

Turn a vague design review into a short, evidence-backed fix list.

Safe working boundary

Review before editing; every suggested fix must point to visible evidence or source lines.

01 · SKILL STACK

A minimal three-role stack

Each role has one job. The personalized builder can adjust the selection for your Agent without changing the decision logic.

01 · Core execution

web-design-guidelines

Runs the structured interface review.

Main limitation
It depends on network access and a live guideline document whose contents can change over time.
First risk
Rule drift or context-free findings
Source evidence
Pinned · b8caa260
Inspect the Skill
02 · Guardrail

context-map

Keeps the review tied to explicit rules and project context.

Main limitation
The workflow is intentionally brief and may miss dynamic dependencies or runtime ownership.
First risk
Incomplete map treated as full scope
Source evidence
Pinned · 318066d2
Inspect the Skill
03 · Evidence

webapp-testing

Captures comparable visual evidence for the final decision.

Main limitation
It assumes Python Playwright and a runnable local application are available.
First risk
Browser actions against the wrong target
Source evidence
Pinned · 0a64e398
Inspect the Skill

02 · ACCEPTANCE

How to know the task is finished

  1. Review the primary page at three representative widths.
  2. Verify focus order, labels, contrast, and the main interaction.
  3. Keep only findings that include reproduction evidence.

03 · STOP CONDITIONS

When to stop

  • The review claims standards compliance without testing.
  • Screenshots are used as proof of behavior.
  • Suggested changes conflict with the product's established design system.

04 · METHOD

How this guide was produced

SkillSignal maps a defined task to three non-overlapping responsibilities, then selects from reviewed catalog metadata. Agent fit, source state, permissions, and published scores are evaluated with deterministic rules. No LLM writes or reorders the recommendation at request time.