nx_html_extract.nx

nx_html_extract.nx -- minimal HTML body-text scraper (state-machine V2). Takes raw HTML bytes (a literotica chapter page, an nhentai gallery info page, a fanfiction.net story, a generic blog post) and emits: - body text (with block tags reduced to newline boundaries) - the first <title>...</title> content PRIORITY: faithful source preservation. Per user spec ("as close to possible to the source") we do NOT normalize whitespace, do NOT collapse newlines into spaces, do NOT re-flow paragraphs. We only: - strip <script>...</script> blocks (executable) - strip <style>...</style> blocks (rendering) - strip <!-- HTML comments --> - strip every other tag's MARKUP, keeping its TEXT CONTENT - decode the common HTML entities Block-level tags (<p>, <br>, <div>, <li>, <h1>..<h6>) cause a newline to be emitted at the open and close, so paragraph structure survives. Per cardinal feedback-build-intelligence-never-strip-features: when we encounter an unfamiliar tag, we strip the tag but KEEP the text content. Silent content stripping is forbidden. nx_safety_envelope: intended_use: "Extract <title> and readable body text from arbitrary HTML, preserving source text bytes verbatim outside of tag/entity markup." sil_target: SIL2 asil_target: QM dal_target: DAL C iec_62304_class: NONE evidence: [no_floating_point, bounded_loops_per_jpl_rule_2, adversarial_html_strips_script_style, entity_decode_table_static, explicit_state_machine_no_ad_hoc_breaks, no_re_flow_or_normalization_of_body_text] hazard_register: [bug-tape-malformed-unclosed-script-tag,