[pullfrog]

Closes #14.

Section Map's per-section indicator was a glass whose fill percentage and glass size came from Difficulty Ladder's v2 payload. That metaphor encoded difficulty as size and could only ever show a fraction. This replaces it with the semantic tier rail the peer's own guide draws: one fixed-size segment per tier, the first current + 1 lit, and an outline on the tier a section plays in full. Difficulty is now shown by how many segments are lit, never by a segment's size.

What changed

• v3 is preferred. When the peer advertises window._ddCapabilities.sectionsSchema === 'difficulty_ladder.sections.v3', Section Map subscribes to difficulty:sections-updated-v3 and renders current_tier / top_tier / max_tier / has_chart_content directly. Feature-detected via the capability marker, never version-pinned — the payload's own schema field is deliberately ignored.
• v2 is a bounded-transition fallback. A peer that predates v3 still gets difficulty:sections-updated; Section Map reconstructs a coarse rail from fillPercentage (topTier = maxTier = maxDifficulty, currentTier = round(fill × topTier)) and discards glassSize entirely. Delete the fallback when the peer's transition policy drops v2.
• Payloads are normalized, not trusted. Entries without finite numeric tiers (v3) or fillPercentage (v2) are dropped; an unreadable container is ignored so the last good state survives, while a well-formed payload with nothing renderable clears the rails. current_tier: null is "unknown" (an unlit rail), not a zero tier.
• Responsive + accessible. A block too narrow for its segments collapses to a single marker; a resize listener invalidates the per-section signatures so the width-dependent form is re-decided. The readable detail Section name · tier X of Y · full detail at Z is the same wording in the block title and aria-label, and only tier-bearing blocks are focus stops — managed in both the full-render and in-place-update paths (so no phantom focus stop when data disappears).
• In-place updates. Rails repaint from the peer's events via a per-section signature; no full _smRender() per event.
• Glass code deleted (_smRenderGlassFilling, .sm-glass, SM_GLASS_SIZES, glassSize handling); state renamed to _smSectionTiers.
• Tests: tier-model/tier-suffix/rail-markup units, responsive collapse, keyboard a11y, late-provider and v2/v3 subscription, malformed-payload, narrow-section, in-place update and resize. node tests/screen.test.js → 44/44.
• Bumps plugin.json to 1.4.0 and updates README.md / CLAUDE.md / the capabilities block.

Standalone behavior

Unchanged: section colors, labels, click-to-seek and wheel nudging all work with no peer installed. The rail is the only dependent feature, gated on the capability marker — absent peer means no rails, never an error.

  ^{｜ View workflow run ｜ via Pullfrog ｜ Using big-pickle (free) ｜ 𝕏}

[pullfrog]

Fixed the check suite failure on PR #18 (CodeFactor: "Complex Method" on screen.js L659–L701).

Diagnosis: the failed check was CodeFactor flagging _smUpdateTierRails for cognitive complexity. Tests and compliance checks all passed locally (node tests/screen.test.js → 44/44).

Fix: split the rail DOM update and the block attribute/title update out of _smUpdateTierRails into _smUpdateBlockRail and _smUpdateBlockAttributes, leaving the loop as signature check → width math → two helper calls. No behavior change; all 44 tests still pass.

Pushed as d324285 to pullfrog/14-difficulty-tier-rail; working tree clean.

  ^{｜ View workflow run ｜ via Pullfrog ｜ Using big-pickle (free) ｜ 𝕏}

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Files Reviewed (5 files)

• screen.js
• tests/screen.test.js
• plugin.json
• README.md
• CLAUDE.md

---

_{Reviewed by efficient · Input: 62K · Output: 13.7K · Cached: 346.1K}