The Agent Skills Directory

Command Execution via Local Scripts: The skill executes specialized Python logic using uv run scripts/runtime_a11y/__main__.py. This pattern is used to run the core matrix engine and rendering tools located within the skill's own directory structure.
External Dependency Management: The skill invokes npm ci within the scripts/runtime_a11y/ directory to prepare its execution environment. This step involves downloading verified packages from standard Node.js registries as part of the skill's setup process.
Network Probing with Security Safeguards: The skill is capable of performing network requests via its runtime harness for accessibility probing. This capability is managed by a specific protocol requiring the --allow-external flag and explicit user confirmation before accessing any non-loopback (external) hosts, which helps prevent unintended network interactions.
Indirect Prompt Injection Surface: The skill ingests external data sources, including assessor findings and prior reports, to build its coverage matrix. To mitigate risks associated with untrusted data, the skill includes a specific requirement to treat all ingested content as data rather than executable instructions, maintaining a clear boundary between the processing logic and external inputs.

accessibility-coverage-matrix