Article Cleanup Workflow (One Article at a Time)¶
Objective¶
Clean one in-use article per cycle, save it in production-ready form for ISR, and leave an explicit editorial trail for BFF migration.
Per-Article Workflow¶
- Select one article from
docs/shared/editorial-cleanup-tracker.yamlwherestatus: pending. - Open the article markdown in
docs/articles/<slug>.md. 2.1 If legacy naming is poor, use replacement flow: - create a new replacement article with a new slug/filename
- keep old article active while replacement is being finalized
- switch collection references to replacement slug once ready
- then mark old one deprecated
- record mapping in
docs/shared/article-migration-ledger.yaml - Apply canonical cleanup:
- metadata normalization (title/subtitle/author/category/tone)
- author selected from
docs/shared/pseudo-authors.yaml - category selected from
docs/shared/editorial-taxonomy.yamlcategorieslist - tags selected from
docs/shared/editorial-taxonomy.yamltagslist - heading structure normalization (
#/##/###) - content rewrite for clarity and consistency
- reference insertion under
## References - hero image and alt text quality check
- Update frontmatter editorial status:
editorial.status: cleanededitorial.cleanedAt: YYYY-MM-DDeditorial.workflowNote: <short-non-sensitive-note>- Update tracker entry for the same slug:
status: cleaned- append an item to
trail[]with date/action/summary. - Validate locally in ISR (target site page and article detail page).
- Commit with message:
Clean article: <slug> (metadata/structure/references/media)
Cleanup Definition of Done¶
- Metadata matches canonical contract.
- Tone is explicitly labeled and consistent in body.
- Body has coherent section hierarchy with no heading jumps.
- References section exists and supports key claims.
- Media links resolve and alt text is descriptive.
- Tracker is updated with a trail record.
- No real identity/credential data is stored in markdown or public tracker fields.
Fast Execution Rule¶
- Work strictly in priority order from currently used app-suite collections.
- Do not batch multiple article rewrites unless explicitly requested.
- Close one article fully before opening the next.