The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using @axe-core/cli and pa11y to perform accessibility scans on user-provided local development URLs. This is a legitimate and expected use of professional developer tools.
[EXTERNAL_DOWNLOADS]: The skill references downloading software and installing packages from trusted and well-known sources, including the npm registry and official organization websites like NV Access. These resources are standard in the web development and accessibility community.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external URLs via automated scanners and interprets the output. While this presents a theoretical surface for indirect prompt injection, the risk is minimal as the input is typically structured technical reports from reputable tools. Mandatory Evidence Chain: 1. Ingestion points: Dev server URL in SKILL.md. 2. Boundary markers: Absent for scanner output. 3. Capability inventory: Subprocess execution for CLI tools in SKILL.md. 4. Sanitization: Relies on structured output from industry-standard scanners.

testing-coach