The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted UI data (web routes and components) and execute shell commands based on the analysis.
Ingestion points: Application routes and UI components provided as input for testing (SKILL.md).
Boundary markers: Uses a structured YAML output format (a11y_compliance) to separate tool results.
Capability inventory: Employs the Bash tool to run external CLI test runners such as Playwright and pa11y.
Sanitization: Relies on external tools (axe-core, Lighthouse) for processing, with no additional sanitization layer defined in the skill instructions.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute shell commands to perform automated accessibility audits.
Evidence: Execution of npx playwright, npx lhci, and npx pa11y-ci to generate audit reports.
[EXTERNAL_DOWNLOADS]: The skill references several external Node.js dependencies required for its operation.
Evidence: Installation commands for @axe-core/playwright, @lhci/cli, pa11y-ci, and vitest-axe.
Context: These are well-known and widely used open-source accessibility testing tools.

qa-a11y-compliance