WORK-638
ID:WORK-638Status:done

Plan runes keep the content their content models drop, then content-unmatched becomes an error

WORK-631 added content-unmatched: validation reports an authored node that no content-model field matches, because the transform drops it. It shipped as a warning because the measurement found that four shipped runes drop real content today, all in the plan plugin. The plan site renders each entity through its rune ({% expand %}), so the loss reaches plan.refrakt.md.

Measured over this repo's plan/ directory (907 files), as of WORK-631:

Priority:highComplexity:moderateMilestone:v0.41.0Source:SPEC-145
main View source

Criteria completion

Criteria completion: 5 of 5 (100%) checked; tracking started on Oct 8, no incremental history yet0%25%50%75%100%Oct 8Oct 11

Tracking started Oct 8 — check back for trends.

Relationships

Implements 1
Related 2
Branches 5
main current done
main doneclaude/v041-plan-runes-keep-content doneclaude/v041-unmatched-content readyclaude/v041-composed-runes-guide readyclaude/v041-project-rune-dir ready

Acceptance Criteria

  • milestone keeps its ## sections, so a milestone's body renders
  • work, bug and decision keep preamble content that is not a paragraph, for example with a catch-all preamble field
  • Rerunning WORK-631's measurement over plan/ reports no content-unmatched finding
  • CONTENT_UNMATCHED_LEVEL in packages/runes/src/lib/index.ts is raised to error, and the docs and tests that state the severity are updated
  • npm run seo:baseline:check and refrakt contracts --check report no drift on either contract copy, or the drift is reviewed and explained

Approach

Fix the content models first, and only then raise the severity. Raising first would fail the build for every project with a plan directory. refrakt plan reads entities structurally, not through the content model, so the CLI is not affected either way.

References

  • WORK-631: the report, and the measurement in its resolution.
  • SPEC-145 D11.

Resolution

Completed: 2026-10-09

Branch: claude/v041-plan-runes-keep-content PR: refrakt-md/refrakt#700

What was done

  • plugins/plan/src/tags/milestone.ts: the goals/notes sequence fields are replaced by one catch-all body field (match: 'any'), the same shape spec uses. The goals list, notes and every ## section render in the body, in authored order. Top-level --- is stripped.
  • plugins/plan/src/tags/{work,bug,decision}.ts: a catch-all intro preamble field after description, and an intro slot (region, omitWhenEmpty, role: body). It keeps blockquotes, lists, fences, tables and --- written before the first ##.
  • plugins/plan/src/config.ts: the Work, Bug and Decision layouts place intro between tags and body.
  • Lumina work.css, bug.css, decision.css: __intro uses the body's type scale. Blockquotes in it are styled like spec's, and a trailing --- is hidden. milestone.css gains the blockquote rule for its body.
  • plan/work/WORK-224: unescaped backslash-escaped backticks that had turned an inline {% codegroup %} into a real, unclosed tag.
  • CONTENT_UNMATCHED_LEVEL is error. These were updated to state the new severity:
    • docs/configuration/reference.md and extend/rune-authoring/content-models.md;
    • the plan rune pages;
    • validate-content.test.ts and composed-character.test.ts.
  • New plugins/plan/test/kept-content.test.ts. emits-equivalence gains intro: 'body' in its sections table; its snapshot is untouched.
  • Contracts (both copies): intro is added to the three layouts.
  • Changeset content-unmatched-error.md (minor) states the behaviour change and the author's fix: add a field, or set rawBody.

Measurement

WORK-631's measurement, rerun over plan/ (908 files):

  • Before: 803 nodes in 200 files: milestone 500/34, work 264/148, bug 37/16, decision 1/1, plus codegroup 1/1 from the malformed WORK-224.
  • After: 0.
  • refrakt validate on main and plan (and plan --deep): 0 findings. plan-site builds clean.

Notes

  • The slot is intro, not preamble. preamble is a header role, and the engine special-cases data-name="preamble"; this content reads as the opening of the body.
  • A preamble that opens with a blockquote before its lead paragraphs puts those paragraphs in intro, not blurb. That keeps the authored order.
  • plan/ used as a site still has 265 pre-existing findings from other ids: literal tag syntax in prose. They are outside this item, and plan/ is not a configured site.