The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the processing of external data by auditing user-provided URLs via the scripts/audit_a11y.py script.
Ingestion points: The tool accepts a target URL through the --url command-line argument, which is then fetched using the requests library.
Boundary markers: The instructions in SKILL.md and SKILL_HE.md provide clear legal notices and disclaimers, stating that the tool's output is for informational purposes and requires review by certified experts.
Capability inventory: The tool performs structural analysis of HTML using BeautifulSoup. It has no capabilities to execute JavaScript or write to the filesystem based on the ingested data.
Sanitization: The script only performs static analysis of tags and attributes; it does not render or evaluate untrusted content from the target URL.
[DYNAMIC_EXECUTION]: The implementation guide in references/widget-implementation.md describes a standard mechanism for applying user preferences via a synchronous bootstrap script to prevent a flash of unstyled content (FOUC). The use of string-templated JavaScript to toggle CSS classes on the <html> element is a legitimate and common pattern in frontend development.
[SAFE]: All external references and dependencies identified in the skill target well-known and trusted sources. This includes official Israeli government domains (gov.il), legal resources (nevo.co.il), industry standards (w3.org), and established developer documentation from Apple, Google, and React Native. The inclusion of these resources is directly related to the skill's primary purpose and does not present a security concern.

israeli-accessibility-compliance