WORK-587 makes brace-structured code addressable. This item makes everything else addressable, and closes D4's dead-end risk before anyone depends on the feature.
SPEC-131 D11 is the case: there are four structural families, not two, and the two this item adds are where every non-code target lives.
Tracking started Sep 24 — check back for trends.
paired is urgent rather than niceIt is the one gap that fails silently. Point auto at a Svelte file anchored on <script> and delimiter counting terminates at the first } closing a brace block at anchor depth — somewhere inside the script body. The slice balances, so WORK-587's self-check passes and the page renders a plausible wrong span. The self-check cannot catch it: the extent is wrong because the wrong family was used, not because the lexer misread anything.
LANG_MAP already ships markdoc, svelte, vue, html and jsx, so this is reachable today by anyone quoting those files.
For section and paired over Markdown the masker must blank fenced blocks and inline code spans — and on a docs site those fences frequently contain the very tokens being counted. site/content/runes/tabs.md:44 has {% tab %} in an inline code span. The pages most worth quoting with paired are exactly the ones full of fenced examples of the tags.
SPEC-131 says plainly: budget this separately from the C-family masker; it is not the same ~40 lines.
{% tabs %} / {% /tabs %}, <section> / </section>. Count depth.{% snippet /%}, <img>, <br>. There is no close, so a paired scan anchored on one would run to EOF. Detect and return one line.until and through are two attributes on purposeD12: until="^}" in code wants the terminator line in; until="^## " in Markdown wants it out. Either default is silently wrong half the time, on the attribute D4 designates as the escape hatch of last resort. Two attributes, each saying what it means.
extent accepts auto (default), dedent, section, and paired; until / through override all fouruntil excludes its matching line and through includes it; neither is inferred from the file or the strategyuntil / through that never matches refuses rather than returning the rest of the file## heading and a {% %} tag that appear inside a fencededent expands tabs at a width taken from the language tableextent="dedent" is covered by tests against an indentation-structured fixtureextent="section" derives its terminator from the anchor's own level, covered by tests on a ### Markdown heading and a TOML tableextent="paired" balances nested same-name tokens, covered by a test extracting an outer {% tabs %} containing inner {% tab %} blockspaired handles asymmetric, symmetric, and self-closing token shapes; a self-closing anchor returns one line rather than scanning to EOFdedent, section, until or through, covered by a test extracting a prose Markdown section containing unbalanced bracketspaired self-checks on its own token stackextent="auto" against a tag-paired file either refuses or is documented as unsupported — never renders a brace-terminated span from a tag-structured sourcesection / paired support rather than a guessed onesnippet and file-ref doc pages document the new attributes, including the fallbacksdedent is the simplest and was the prototype's first attempt at a universal rule — it failed there (a multi-line parameter list returns to the anchor's indent before the body opens, truncating roughly one TypeScript declaration in eight) and works well as a declared strategy. Do not be tempted to promote it.
Never infer strategy from the file (D5). Guessing dedent for .py and auto for .ts would be right most of the time and silently wrong the rest, reintroducing the failure mode the spec exists to remove. Table lookups for lexical facts; never for strategy.
Order within the item: until / through first (they are small and they are what every refusal message points at), then dedent, then the Markdown masker, then section and paired on top of it.
WORK-590's migration should not start before this lands — section and paired are what make the non-TypeScript half of site/content addressable at all.
until vs through)site/content/runes/tabs.md — the inline-code-span case the Markdown masker must surviveCompleted: 2026-09-24
Branch: claude/v0-37-0-review-vqpl41 PR: refrakt-md/refrakt#647 (batched with WORK-587 and WORK-589)
dedent, section, paired and the explicit until / through terminators, in lib/anchor.ts.lib/mask.ts — its own pass, selected by a language declaring fences / inlineCode rather than by name.delimited added to the language table.site/content/runes/snippet.md and file-ref.md.delimited is the mechanism behind two separate criteria. It is a lexical fact — "does a ; or a matching } end a construct here" — not a strategy, so D5 holds: the engine still never guesses which extent to use. It lets auto refuse in Python and YAML instead of terminating on the next line, and it is what makes auto against a tag-paired file refuse rather than render a brace-terminated span from tag-structured source.paired close. Worth preserving if this code is ever revisited.section and paired read the masked lines, not the raw ones. That is the whole point of the Markdown pass — reading raw was the bug.## inside a fence, {% %} inside an inline span) failed immediately. Unchecked, implemented, re-checked. WORK-597's test asserting the masker ignores Markdown was updated rather than deleted, since it was correct when written.paired self-checks on its own token stack: an unbalanced scan returns undefined and refuses. There is no separate delimiter balance for it, which is correct — D2 scopes that to auto.4820 tests pass, 0 failures. Site content validates clean.