The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because its primary function involves the ingestion and analysis of external web content. Maliciously crafted data on an audited page could attempt to influence the agent's behavior or reporting during the analysis phase.\n- Ingestion points: External content is fetched using Puppeteer via page.goto(url) in the AccessibilityAuditor class located in resources/implementation-playbook.md.\n- Boundary markers: The skill does not currently specify the use of delimiters or 'ignore' instructions for the web content being processed.\n- Capability inventory: The skill utilizes Puppeteer for browser automation and is capable of performing network requests and report generation.\n- Sanitization: The implementation focuses on accessibility metadata extraction and does not include mechanisms to sanitize or filter natural language prompts that might be present in the audited HTML.

accessibility-compliance-accessibility-audit