audit agent stuck on modal dialog during axe scan

audit agent stuck on modal dialog during axe scan - how to fix it

TL;DR

Teach the agent a modal protocol: detect the dialog, either dismiss it (Escape, close button) or audit inside it, then continue. An agent that treats a modal as a dead end stalls forever. One line of why: modals trap focus and cover the page, so the agent's normal interact-and-scan loop cannot proceed until the modal is handled.

The error, verbatim

AuditAgentError: agent stuck on modal dialog, no progress for 120s
    dialog: .newsletter-modal (role=dialog, aria-modal=true)
    last_action: click body (no-op, modal overlay intercepts)

Fix it step by step

Step 1: Reproduce the stall

node agent/run-audit.js --route /home | rg -i 'stuck|modal' | head -5

Expected: The agent stalls when the newsletter modal appears.

Step 2: Confirm the modal traps focus

npx @axe-core/cli https://example.com/home --rules aria-hidden-focus | tail -3

Expected: Shows whether the modal is at least implemented correctly; a broken modal needs a product fix too.

Step 3: Add the modal protocol

rg -n 'modal|dialog' agent/actions.js | head -10

Expected: Find the agent's action loop to add: detect role dialog, try Escape, then the close button, then audit-inside.

Step 4: Re-run the audit

node agent/run-audit.js --route /home | tail -4

Expected: Agent dismisses or audits the modal and completes the page scan.

Step 5: Add a regression probe

node agent/run-audit.js --smoke | tail -3

Expected: Smoke run passes; schedule it so the breakdown is caught if it ever regresses.

When to use this skill

• You run an agent that scans UIs for accessibility and it hits this breakdown
• The agent's scan loop stalls, crashes, or loops on this exact failure
• You are hardening an audit agent's error handling for production scans

When NOT to use this skill

• A human runs the scan manually and it works, this is agent-harness failure handling
• The scan completes and only reports violations, use the rule-specific skills

Compatibility

Audit agent harness with DOM action primitives (click, press Escape, query role=dialog). Works with Playwright or Puppeteer drivers. Pin the tool version in the lockfile so scans stay reproducible across machines.

Variant phrasings

agent stuck on popup during scan

Same stall, popups and modals alike.

axe scan blocked by modal

Practitioner phrasing for the same protocol.

the breakdown hits other routes too

Agent failure modes are systemic; apply the hardening to every route the agent covers, not just the one that failed.

Why it happens

Marketing modals (newsletter, discount) appear on timers or scroll depth, right in the middle of an agent's scan. The agent's click actions hit the overlay and no-op, its scroll does nothing, and with no modal concept it retries forever. The fix is a small state machine: on detecting an open dialog, decide dismiss vs audit-inside, execute, verify the dialog closed or the inner audit completed, then resume. Dismissal should prefer Escape, then a labeled close button, and never click randomly. Agent breakdowns are systemic: the same failure mode will hit every route, page, or run the agent touches. Harden the harness once (timeouts, loop detection, verification gates) instead of patching per page, and keep breakdown telemetry separate from violation counts.

Edge cases

• Some modals are the audit target (checkout dialogs), audit inside them instead of dismissing.
• Cookie-consent dialogs need an accept or reject decision, record which one the agent chose.
• If the modal reappears every navigation, the agent needs a seen-set so it does not loop dismissals.
• Log breakdowns separately from violations in agent telemetry; mixing them hides whether the agent itself is getting more reliable.

Provenance

Resolved from the public thread: https://vectle.com/posts/pstuwitbKzdzPoIajY_bdccg