transformRune still strips consumed author attributes from pass-through output using a fixed list naming eight facet-owned axes:
c2c17097dee74d5a34e38Facet gains a static attributes?: readonly string[] declaration — the author attributes that facet ownswidth, spacing, inset, elevation, prominence, motion (reveal + stagger) and density each declare their ownattributes, plus data-rune / data-rune-fields, plus the config-modifier facet's dynamic stripAttrstransformRune is gone, along with the NOTE comment marking itnpm run build, the full repo suite, and refrakt contracts --check passThe reason this needs a static declaration rather than the existing FacetResult.stripAttrs channel is subtle and worth stating: these attributes must be stripped even when the facet resolves to nothing. width="content" is suppressed (the facet returns null) but the attribute must still not reach the output. A per-instance result channel cannot express that; stripAttrs stays for the config-modifier facet, whose attribute names come from config and so cannot be static.
Deliberately split out of WORK-523 rather than folded into it — that item was the riskiest migration in the milestone (every rune goes through the modifier loop), and widening it for a mechanical cleanup would have been a poor trade.
stripAttrs and left this residueCompleted: 2026-09-02
Branch: claude/transform-package-refactor-7mxi8a
packages/transform/src/facets/types.ts — Facet.attributes, a static declaration of the author attribute names a facet owns.packages/transform/src/facets/{box,elevation,prominence,motion,density}.ts — each declares its own: width, spacing, inset, elevation, prominence, reveal + stagger, density.packages/transform/src/facets/index.ts — FACET_ATTRIBUTES, the union derived once at module load.packages/transform/src/engine.ts — the fixed destructure and its NOTE comment are gone; pass-through now filters on FACET_ATTRIBUTES plus the config-modifier facet's dynamic stripAttrs, with only data-rune / data-rune-fields named directly.packages/transform/test/facets/registry.test.ts — 10 tests covering the derived set, registry ordering invariants (every facet after its declared dependencies, no duplicate names), and stripping including the resolved-to-nothing case.Static, not a result field. These attributes must be stripped even when the facet resolves to nothing: width="content" is suppressed — no axis, no class — but must not reach the rendered element. A per-instance channel cannot express that, which is why stripAttrs alone was insufficient and why this was split out of WORK-523 rather than folded in.
An asymmetry found and deliberately preserved. Checking the pre-change behaviour attribute-by-attribute before touching anything showed that reading and dropcap reach the rendered element today, unlike the other eight. Both are consumed and re-expressed as data-reading / data-dropcap on the body section, so the raw attributes surviving looks like an oversight — but removing them is a behaviour change, and this item is a mechanical cleanup. They are excluded from FACET_ATTRIBUTES and pinned by a test recording why, so the exclusion does not read as arbitrary. Worth its own item if it should be fixed.
A flake, unidentified. The first full-suite run reported 1 failure; it cleared before I captured which test, and three subsequent full runs were green at 4012. The same container produced a plugins/plan timeout flake earlier in this milestone, so it is plausibly the same timing sensitivity — but I cannot name it, so it is recorded rather than dismissed.
Behaviour verified attribute-by-attribute against the pre-change baseline, including an unrecognised attribute still passing through untouched. All 630 pre-existing transform tests pass unmodified.