remediation agent stuck in loop fixing heading order

remediation agent stuck in loop fixing heading order - how to fix it

TL;DR

Fix the outline in one pass from a single scan: collect every heading violation, compute the corrected levels for the whole tree, and apply them together. Agents that fix one heading, re-scan, and fix the next loop because renumbering shifts the outline. One line of why: heading levels are relative to each other, so local fixes invalidate each other.

The error, verbatim

RemediationAgentError: heading order loop, 15 iterations
    pattern: fix h4 to h3, rescan finds new skip at h5, repeat
    page: docs with component-injected headings

Fix it step by step

Step 1: Reproduce the loop

node agent/remediate.js --rule heading-order | rg -i 'loop|iteration' | head -5

Expected: The agent renumbers headings one at a time forever.

Step 2: Dump the full outline

npx @axe-core/cli https://example.com/docs --rules heading-order --save axe-ho.json | tail -2

Expected: One scan shows every offending heading at once.

Step 3: Renumber in one pass

rg -n 'heading' agent/patches/heading-order.js | head -10

Expected: Change the patch: compute all corrected levels from one outline, apply together.

Step 4: Re-run remediation

node agent/remediate.js --rule heading-order | tail -4

Expected: Single pass produces a valid outline, verification passes.

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

Remediation agent with DOM patch primitives. One axe scan provides the full heading outline. Pin the tool version in the lockfile so scans stay reproducible across machines.

Variant phrasings

agent heading renumber loop

Same loop, same single-pass fix.

fix agent heading order infinite

Practitioner phrasing.

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

Heading levels are globally relative: changing one heading's level changes what its neighbors should be. An agent fixing violations one at a time chases a shifting outline, especially when components inject headings at render. The correct patch takes the whole outline from one scan, computes a valid leveling (each increase steps by one), and applies all changes atomically. Component-injected headings may need the component fixed instead, if the outline shifts per render. 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

• Components that render headings need a level prop, otherwise the outline depends on page context.
• Decreases can skip levels freely, only increases must step by one, the algorithm should know this.
• Verify the outline after, not just the rule: a valid outline is the real goal.
• 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/pst_ewqNRS4h1Z5-8D-gHm-nIA