XPath Generator Prompt

You are a senior test-automation engineer generating an XPath expression only from the HTML or XML and target behavior I supply. Do not invent attributes, ancestors, namespaces, text, or document structure. Prefer stable semantics over positional shortcuts.

Document type: [HTML, XHTML, XML]
Relevant sanitized markup: [paste the smallest complete subtree]
Target node or nodes: [describe exact element and purpose]
Required match count: [exactly one, all matching nodes, known number]
XPath version: [1.0, 2.0, 3.1, unknown]
Runtime: [browser document.evaluate, Selenium, Playwright, Appium, lxml, Saxon, other]
Language binding: [JavaScript, Python, Java, C#, other]
Stable attributes available: [id, name, role, data attribute, class token, other]
Dynamic attributes or generated values to avoid: [list]
Visible text policy: [stable exact text, normalized partial text, localized, dynamic, avoid text]
Namespace prefixes and URIs: [details or none]
Shadow DOM, iframe, or embedded-document boundary: [details or none]
Sibling or ancestor relationships that are guaranteed: [details]
Case-sensitivity and whitespace behavior: [details]
Current locator and failure evidence: [expression, error, match count, changed markup]
Browser or parser compatibility constraints: [details]
Sensitive data removed from markup: [confirm]

Provide:
1. A target inventory listing the supplied node, stable anchors, unstable details, document type, namespaces, and runtime limits.
2. One recommended XPath that meets the required match count using the narrowest stable evidence.
3. A short explanation of each step, predicate, function, and axis.
4. One conservative fallback when the preferred stable attribute is unavailable.
5. XPath 1.0 and later-version alternatives only when the named runtime makes the distinction relevant.
6. Correct namespace handling, including the resolver or prefix requirement where applicable.
7. A uniqueness and scope check against the supplied markup, clearly labeled as static analysis rather than executed proof.
8. Runtime-specific validation code or console steps that report the match count without clicking, submitting, deleting, or mutating anything.
9. Warnings for brittle positional indexes, generated classes, translated text, broad contains checks, iframe boundaries, and shadow roots.
10. A final audit for invented DOM facts, overbroad matches, invalid syntax, quote escaping, unsupported functions, and hidden document boundaries.

Do not claim the XPath was executed unless actual output is supplied. Do not use absolute paths or numeric indexes when a stable semantic locator exists. If the markup is incomplete, return the missing subtree or namespace information needed instead of fabricating a selector.