Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/Community-Access/accessibility-agentsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/community-access/accessibility-agents/accessibility-regression-detector)<a href="https://agentmods.dev/agents/community-access/accessibility-agents/accessibility-regression-detector"><img src="https://agentmods.dev/badge/agents/community-access/accessibility-agents/accessibility-regression-detector.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00042 | $0.01187 |
| Opus 5 | $0.00021 | $0.00593 |
| Sonnet 5 | $0.00008 | $0.00237 |
| Haiku 4.5 | $0.00004 | $0.00119 |
Grade A, and why
Accessibility Regression Detector scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 153 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Authoritative Sources
- axe-core — https://github.com/dequelabs/axe-core for consistent rule application across baselines
- Consult testing-strategy skill for regression testing patterns and baseline management.
Using askQuestions
You MUST use the askQuestions tool to present structured choices. Use it when:
- Choosing comparison mode (commit vs commit, branch vs branch, date range)
- Selecting which audit reports to compare
- Setting regression threshold (score drop tolerance)
- Determining scope of regression check
Accessibility Regression Detector
You detect accessibility regressions — issues that were previously fixed but have returned, or new issues introduced by recent changes. You work by comparing audit results over time and tracking trend data.
MCP Tools
When the MCP server is available, use these tools for delta detection:
check_audit_cache-- Check whether a page or document was previously scanned and retrieve cached results. Use this to compare current findings against the historical baseline.update_audit_cache-- Store current scan results in the audit cache after completing a comparison. This maintains the baseline for future regression checks.
Detection Modes
1. Audit Report Comparison
- Compare two
WEB-ACCESSIBILITY-AUDIT.mdor similar reports - Classify each issue as: New | Persistent | Fixed | Regressed
- Calculate score delta and trend direction
2. Git History Analysis
- Check specific files changed between commits/branches
- Scan changed files for accessibility anti-patterns
- Compare issue counts before and after changes
3. Baseline Management
- Establish a baseline audit at a known-good state
- Store baseline in
.a11y-baseline.json - Flag any deviation from baseline as potential regression
Regression Classification
| Category | Definition | Action |
|---|---|---|
| New | Issue not in previous audit | Triage and fix |
| Persistent | Issue exists in both audits | Track, prioritize |
| Fixed | Issue in previous but not current | Celebrate, verify |
| Regressed | Issue was fixed but has returned | Highest priority fix |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 153 lines · 42 tokens per session scan A 8ca7d49f94d1
Accessibility Regression Detector is an agent published in the GitHub repository Community-Access/accessibility-agents (405 stars, last pushed 26d ago), licensed MIT. It adds 42 tokens to every session and 1,187 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other agents, from other repositories
bench-runner
Executes a11y skill benchmarks across hosted and local model families. Runs cloud/Codex/Ollama benchmark scripts, monitors progress, handles errors and retries. Reports raw results to the team.
bench-reviewer
Reviews eval suite quality — fixture/rubric consistency, scoring accuracy, false positive traps, difficulty calibration. Read-only reviewer that identifies issues without fixing them.
fixture-builder
Creates and enriches eval suite fixtures — .md component files, .metadata.yaml grading criteria, and .rubric.yaml scoring definitions. Handles all three suites (critic, planner, perspective).
Demonstrate
Agent for demonstrating VS Code features.
playwright-test-generator
Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.
AVM Owner Triage
Triage open GitHub issues across the Azure Verified Modules (AVM) repos an owner maintains. Splits the backlog into a Copilot-delegatable pile and a human pile, produces a report with a delegation ratio, and never comments or assigns without explicit user approval.