Markdown work checklist

Updated 2026-10-23. Ordered by current relevance. Checked means the named work is complete within its stated boundary; it does not imply deployment or full Markdown compatibility. Optional syntax entries are decisions to implement/defer/omit, not approved implementation requirements. Keep historical evidence intact and update this checklist when work changes.

1. Current task-checkbox acceptance — highest priority

• Fix barely distinguishable checked/unchecked appearance with hollow outlines versus solid fills and contrasting ticks; retain native disabled checkbox semantics, accessible task names and working links. Initial native-paint acceptance failed user UAT and was corrected.

• Verify corrected appearance in Chromium across all 16 registered themes, phone/desktop, Auto light/dark, LCD palettes and forced colors: 46 dynamic + 46 generated cases; inspect all state-pair screenshots. Physical-device/AT acceptance remains below.

• Complete user UAT of the corrected checkbox slice: user accepted it and authorized merge on 2026-10-23.

• Merge the accepted task-checkbox slice into main. Production deployment/refresh remains a separate item below.

• Implement checked/unchecked display for supported unordered task items, including uppercase X, wrapped task labels and nested items.

• Keep task display immutable: disabled, skipped by Tab, no source writes, submission or editing controls.

• Preserve literal markers in ordered items, code/table examples, malformed/empty labels and unsupported whole-list source blocks.

• Verify task names/links, source downloads, dynamic/static parity and publication failure/recovery. Initial corpus: 1,109 displayed controls and 73 retained source candidates; corrected appearance has separate verification below.

Evidence: task implementation, task corpus/acceptance, contrast correction evidence.

2. Explicitly requested usability follow-ups

• Implement an in-page Markdown checkbox counter showing checked/total; decide whether the count covers rendered controls or other source markers during its own planning.

• Relax the narrow Docs column on wider screens to reduce avoidable table/deep-list scrolling; retain phone containment and keyboard scrolling when content really is wide.

• Record the counter in todo.txt and the wider-column request in the renderer handoff.

These are separate future slices, not part of the current checkbox correction.

3. Core renderer work already delivered

• Restore the project-owned dependency-free renderer; remove the eight-package Markdown dependency bundle and its runtime imports.

• Preserve headings, duplicate-safe heading IDs/fragments, ordinary emphasis, inline code, safe inline links, quotes and thematic breaks.

• Render standalone pipe tables: header/body cells, alignment, optional outer pipes, empty cells, escaped pipes and supported inline-code pipes.

• Provide keyboard-accessible table scrolling and whole-block source fallback for uncertain tables.

• Preserve ordered/unordered nesting, mixed child types and ordinary continuation/loose paragraphs within the supported indentation rules, through eight levels.

• Preserve explicit source step numbers, including starts, repeats, restarts and gaps.

• Keep unsupported/ambiguous list blocks readable as source, without consuming following standalone content.

• Retain fenced example bytes, language captions, keyboard source/code regions and readable Mermaid source.

• Retain collapsed escaped metadata, hidden prose comments, escaped HTML, image alternatives/no fetching and traversal/symlink/unsafe-link protections.

• Deliver the same renderer through ordinary PHP Docs and full/shared static publication, with stale refresh, failure retention and restoration/retry evidence.

• Freeze source metrics, classify measured tables/lists/tasks and retain per-document hashes without restoring a reference-parser dependency.

• Keep Docs discovery, browsing/search, safe downloads and root-document reading integrated with the existing reader.

Evidence: table slice, list slice, renderer runbook.

4. Optional compatibility decisions — prioritize by real document needs

• Decide reference-style links/definitions: explicit, collapsed and shortcut forms, preserving document-boundary safety.

• Decide URL autolinks: bare URLs, angle-bracket URLs and email autolinks.

• Decide underscore emphasis/strong emphasis and strikethrough.

• Decide hard line breaks from trailing spaces or backslashes.

• Decide Setext headings, grouped/nested blockquotes and other currently unsupported block forms.

• Decide tilde fences, standalone indented code and more complete inline-code delimiter/escape handling.

• Decide broader list support: embedded tables/quotes/fenced or indented code, tabs/other indentation conventions and nesting beyond the current bound.

• Decide task presentation in ordered lists; retain numbering and read-only behavior if selected.

• Decide richer table forms: tables in lists/quotes, multiline/merged cells, nested tables, uneven rows and more complex code-span boundaries.

• Decide rare or currently unused extensions such as footnotes and admonition-style alerts.

• Decide whether Mermaid/math rendering or actual image loading is wanted; each needs a separate scope/security/dependency review. Labeled source/text alternatives remain the current behavior.

• Reconcile the historical rare-syntax decision table with completed slices and record explicit implement/defer/omit decisions, including zero-use variants; do not replace historical counts with invented current measurements.

Evidence: historical rare-syntax table, original formatting inventory, latest task-slice metrics. Candidate counts are heuristics, not conformance guarantees.

5. Verification and documentation closeout

• Correct runbook verification wording to expect supported rendered tables and keyboard table/source/code regions.

• Complete physical-device and assistive-technology acceptance for the recent renderer slices; browser accessibility checks are not a substitute.

• Run/classify an appropriate current full-suite result before claiming full-suite verification; preserve previously classified unrelated baseline failures.

• Deploy the accepted renderer to the intended production instance, refresh generated Docs and verify live reading/recovery there.

• Complete focused renderer/catalog/static checks, copied-package recovery and representative dynamic/static phone/theme scenarios for table, list and initial task slices; checkbox painting was corrected and locally reverified above; user UAT remains pending.

• Preserve the rule requiring explicit approval for any new external dependency; none was added by these extension slices.

6. Original dependency/history-removal closeout — separate from renderer features

• Publish the reviewed rewritten main/refs and verify a fresh independent clone contains none of the targeted paths/blobs.

• Synchronize the seven inventoried clones, retained refs, eight stashes and fourteen worktree registrations while preserving local work/configuration and the publication audit.

• Review and explicitly authorize any disposal of retained old objects, non-stash reflogs, stale worktree registrations or private recovery bundles. They remain intentionally preserved.

• Account for other deployment, release/archive and external copies, plus host-retained objects/caches; current evidence does not establish universal erasure.

• Record final maintainer acceptance of the full removal outcome and its retention limits.

Evidence: removal implementation, clone/history handoff.

Current policy, not unfinished implementation

Raw HTML stays escaped, arbitrary images are not fetched, Mermaid stays source, and checkboxes remain read-only. Changing these policies is not implied by an unchecked optional decision. Do not delete recovery data or modify todo.txt without the required explicit user authorization.