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.
npx agentmods add instructions/elizabeth1979/accessibility-engine/claude-mdgit clone --depth 1 https://github.com/Elizabeth1979/accessibility-engineWrote 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/instructions/elizabeth1979/accessibility-engine/claude-md)<a href="https://agentmods.dev/instructions/elizabeth1979/accessibility-engine/claude-md"><img src="https://agentmods.dev/badge/instructions/elizabeth1979/accessibility-engine/claude-md.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 | $0.00533 | $0.00533 |
| Opus 5 | $0.00267 | $0.00267 |
| Sonnet 5 | $0.00107 | $0.00107 |
| Haiku 4.5 | $0.00053 | $0.00053 |
Grade A, and why
accessibility-engine CLAUDE.md 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 — 21 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Accessibility Evidence Engine — agent guide
AI-first accessibility testing: AI elevates each check from "is the attribute present?" to "is it correct in context, and here's a better value." Implemented end to end across Tiers 1–5 (naming, vision, dynamic, the axe-core floor, advisory) and verified on a local model; the engine, the agent surfaces (MCP, triage), and remediation (fix) are real. See docs/ROADMAP.md for current state and what's next, and docs/PLAN.md for the architecture.
Frozen invariants (do not break)
- AI sees captured evidence only, never the live page.
@aee/aidepends on@aee/coreonly.tests/graph-guard.test.jsenforces this and the other graph edges — keep it green. The only package allowed to depend on both a driver (@aee/playwright) and the judges is@aee/engine. - Verdicts never silently upgrade: UNKNOWN and advisory results must never become PASS.
Defaults
- Local-first, no API key. The AI layer is provider-agnostic (the
JudgmentModelseam). For local runs setAEE_LLM_PROVIDER=local(Ollamagemma4:e4bby default; see README).claudeandstubproviders also exist; with no key the default isstub, and a local judgment degrades to UNKNOWN if the server is down — never a guessed PASS. - Tests:
pnpm testrunsnode --testover builtdist/. Each change leaves a green build. Browser and local-model tests are gated (they skip when a browser / local server is unavailable). - Runs are in-memory by default. Set
AEE_STORE_DIRto persist runs + their content-addressed artifacts to disk so they survive across processes; the engine still resolves artifact refs before the AI (the evidence-only invariant holds either way).
Conventions
- TypeScript ESM, NodeNext resolution,
.jsimport extensions,zodas the single schema source of truth (types viaz.infer). - Plans and docs: clean reflowable markdown (no hard-wrapped "bricks", no changelog sections), committed in-repo.
- Commit and push only when asked. The repo has a GitHub remote (
origin→ github.com/Elizabeth1979/accessibility-engine, public); commits land onmain.
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 · 21 lines · 533 tokens per session scan A c86321e4b2b1
accessibility-engine CLAUDE.md is an instructions file published in the GitHub repository Elizabeth1979/accessibility-engine (0 stars, last pushed 2mo ago), licensed MIT. It adds 533 tokens to every session, about $0.0027 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 2026-08-31.
Other instructions, from other repositories
ios-swiftui-accessibility-techniques CLAUDE.md
Claude Code instructions for cvs-health/ios-swiftui-accessibility-techniques, covering project rules, copyright header (required), rules, swift files and markdown files.
ACCESSIBILITY.md AGENTS.md
AGENTS.md instructions for mgifford/ACCESSIBILITY.md, covering repository instructions for ai coding agents, scope, when each instruction file is useful, project mission and status and required project context.
monochrome-ui AGENTS.md
AGENTS.md instructions for vaneenige/monochrome-ui, covering agents.md, north stars (non-negotiables), why the core looks weird (and should stay weird), clever tricks and code style.
ACCESSIBILITY.md copilot-instructions.md
Copilot instructions for mgifford/ACCESSIBILITY.md, covering github copilot repository instructions, purpose and usefulness, required behavior, copilot-specific notes and where new copilot guidance belongs.
kolibri copilot-instructions.md
Copilot instructions for public-ui/kolibri, covering github copilot instructions for kolibri repository, kolibri overview, essential setup commands, 1. install node.js 22 and 2. enable pnpm.
arena AGENTS.md
AGENTS.md instructions for dravensoft-dev/arena, covering arena, for whoever changes it, which job is this?, where each decision goes, what this repo ships and where a new document goes.