Awesome Reviewers

When parsing configuration or design-token files from external/untrusted sources, never execute their contents as part of scanning. Default to static, deterministic extraction and “fail closed” (e.g., return null/empty) when the config requires evaluation.

Apply this pattern:

• Default non-execution: Use static parsing (e.g., balanced-brace/quote-aware extraction of literals) instead of require, import, function/arrow evaluation, template literals, spreads, or any construct that implies runtime behavior.
• Deny dynamic constructs: If the parser encounters unsafe indicators (like require(, import(, =>, function, template literals, process., new, etc.), disqualify and return “unknown” rather than guessing.
• Require explicit opt-in for execution: If execution is unavoidable, gate it behind an allow_*_execution flag that is off by default, clearly documented as dangerous, and tracked in metadata/warnings.
• Prove with tests: Add tests that assert the execution path (e.g., subprocess.run, node) is never called in the default path.

Example shape (self-contained):

def parse_tailwind_config(path: str, allow_config_execution: bool = False):
    text = open(path, 'r', encoding='utf-8', errors='replace').read()

    # Fail-closed on dynamic constructs
    unsafe_markers = [
        'require(', 'import(', '=>', 'function', 'new ', 'process.',
        '`', '{...', '...'
    ]
    if any(m in text for m in unsafe_markers):
        if not allow_config_execution:
            return None  # refuse
        # If opt-in is enabled, execution is explicit and auditable.
        # (Keep this codepath tightly scoped and test-covered.)
        return execute_config_safely(path)

    # Otherwise extract literal theme statically (quote-aware / balanced braces)
    return extract_theme_literal_as_json(text)

This prevents supply-chain style attacks where a config triggers arbitrary code execution during analysis, aligning scanning tools with security best practices.

This site is open source. Improve this page.