The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute local commands for its core functionality: using a Python script (scripts/inline-components.py) to bundle CSS into the artifact and xdg-open to display the final result. These commands operate on locally generated files and are consistent with the tool's intended purpose.
[PROMPT_INJECTION]: The skill handles untrusted data (source code, PR bodies, datasets) which presents an indirect prompt injection surface. This risk is effectively mitigated by the skill's own strict guidelines in references/quality-and-validation.md, which mandate JSON serialization, escaping of HTML special characters (e.g., < to \u003c), and the use of safe rendering methods like textContent instead of innerHTML.
[DATA_EXFILTRATION]: No data exfiltration patterns were found. The skill explicitly requires artifacts to be offline documents with no CDN dependencies, fetch calls, or remote resource loading, ensuring that processed data remains local to the user's environment.
[CREDENTIALS_UNSAFE]: The skill includes explicit instructions for the agent to exclude secrets, tokens, and private environment variables from the generated output, promoting safe handling of sensitive information.

html-artifacts