Accessibility Auditor
v1.0.0Expert AI agent specializing in accessibility auditor. From The Agency (github.com/msitarzewski/agency-agents).
⭐ 0· 76·0 current·0 all-time
by@zhouqkt
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
medium confidencePurpose & Capability
The name/description match the instructions: auditing against WCAG, running automated scans (axe-core, Lighthouse) and manual assistive-technology testing. However, the SKILL.md assumes access to runtime tools (npx/node, browsers, screen readers) while the declared requirements list no required binaries — that mismatch should be resolved or documented.
Instruction Scope
Instructions are narrowly focused on accessibility auditing tasks (automated scans, manual testing, reporting, remediation). They do not direct the agent to read unrelated files, exfiltrate secrets, or contact opaque endpoints. The guidance does require manual testing steps that imply human or environment capabilities (screen readers, OS/browser access).
Install Mechanism
This is an instruction-only skill with no install spec or code files. That minimizes installation risk. Note: runtime commands (npx) will fetch packages on demand if executed, but there is no install-time download specified in the skill bundle.
Credentials
The skill requests no credentials or config paths, which is appropriate. It implicitly requires Node/npm (npx) and access to browsers and assistive-technology tools for manual testing; these needs are not declared as required binaries. No secrets or unrelated environment variables are requested.
Persistence & Privilege
The skill does not request always-on presence and uses default invocation settings. It neither alters other skills nor demands system-wide configuration changes.
Assessment
This skill appears to do what it claims (accessibility audits), but before installing consider: (1) it assumes you can run tools like npx, Lighthouse, and axe-core — ensure Node/npm is available or the agent will fail when trying 'npx' commands, (2) many checks require human/manual testing with screen readers, browsers, zoom, and OS-specific tools (VoiceOver, NVDA, JAWS); an autonomous agent cannot fully perform those steps without human-in-the-loop or a suitably provisioned environment, (3) because the skill is instruction-only it doesn't install code itself, but running the suggested commands will cause npx to download packages at runtime, so only allow execution in a trusted environment, and (4) if you want stronger guarantees, ask the publisher to add explicit required-binaries metadata (e.g., node/npm, browsers) and clarify which steps need human testers. If you don't want the agent to run shell commands on your machine, avoid granting it execution permissions or run the audit steps manually.Like a lobster shell, security has layers — review code before you run it.
agencyvk9721q1jygy1v6w8k0rh9qdfcx84kvx5ai-agentsvk9721q1jygy1v6w8k0rh9qdfcx84kvx5latestvk9721q1jygy1v6w8k0rh9qdfcx84kvx5
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
Runtime requirements
🤖 Clawdis