The Agent Skills Directory

[COMMAND_EXECUTION]: The skill contains a suite of Python scripts in the scripts/ directory designed for format conversion. Scripts such as convert_to_json.py, convert_to_csv.py, and batch_convert_templates.py use subprocess.call to execute a local conversion utility (convert_formats.py). These operations are restricted to the skill's local environment and are used to manage test artifacts.
[COMMAND_EXECUTION]: Installation instructions in README.md suggest running a PowerShell script with -ExecutionPolicy Bypass. While this is a common method for installing tools on Windows systems, it involves executing script logic that bypasses default system restrictions.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external product documentation, requirements, and design notes. This ingestion of untrusted data from the user's project context creates a potential attack surface for indirect prompt injection.
[INDIRECT_PROMPT_INJECTION]: Evidence Chain:
Ingestion points: Documentation and design materials are ingested via the main prompt defined in prompts/accessibility-testing.md.
Boundary markers: No specific boundary markers or isolation instructions are employed to separate user-provided data from the system's operational instructions.
Capability inventory: The skill utilizes file system operations and local script execution through subprocess.call to process and export data.
Sanitization: The skill does not implement automated sanitization or filtering for the external content it processes.

accessibility-testing