The Agent Skills Directory

[COMMAND_EXECUTION]: The skill references standard accessibility auditing tools intended for diagnostic purposes in references/testing-triage.md.
The instructions suggest using lighthouse, axe, and timeout to perform automated audits on target URLs.
These are well-known, legitimate tools for web development and accessibility testing.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for auditing external or user-provided web content, creating a potential surface for indirect injection.
Ingestion points: The agent ingests HTML and web content during audits or remediation tasks as described in SKILL.md.
Boundary markers: None explicitly defined in the prompt instructions.
Capability inventory: The skill includes the ability to execute shell commands (lighthouse, axe) and modify source code.
Sanitization: The diagnostic tools used are automated scanners, which typically do not interpret natural language instructions from the scanned content, significantly mitigating the risk.
[EXTERNAL_DOWNLOADS]: The skill references several external websites and a GitHub repository for documentation and licensing.
References include the W3C normative standards (w3.org), Deque University accessibility rules (dequeuniversity.com), and Google's developer documentation (web.dev).
A link to the original source repository addyosmani/web-quality-skills on GitHub is provided in NOTICE.md for license compliance.

z-accessibility