This release includes breaking changes for platform teams planning a safe upgrade.
✓ No known CVEs patched in this version
Summary
AI summaryUnified colour‑blind‑safe doc‑navigability graph replaces docs‑staleness treemap.
Changes in this release
| Type | Severity | Summary | CVE |
|---|---|---|---|
| Feature | Low |
Adds unified doc-navigability graph for /assess, colour‑blind‑safe by default. Adds unified doc-navigability graph for /assess, colour‑blind‑safe by default. Source: llm_adapter@2026-05-28 Confidence: high |
— |
| Feature | Low |
Implements Lint health‑check scanning hubs, orphans, connectivity, dangling references, and missing cross‑references. Implements Lint health‑check scanning hubs, orphans, connectivity, dangling references, and missing cross‑references. Source: granite4.1:30b@2026-05-28-audit Confidence: low |
— |
| Feature | Low |
Updates Claude Code plugin and standalone Desktop/Cowork ZIPs to latest rolling release. Updates Claude Code plugin and standalone Desktop/Cowork ZIPs to latest rolling release. Source: granite4.1:30b@2026-05-28-audit Confidence: low |
— |
| Performance | Low |
Scans only git‑tracked repository files; ignores untracked personal files and external symlinks. Scans only git‑tracked repository files; ignores untracked personal files and external symlinks. Source: granite4.1:30b@2026-05-28-audit Confidence: low |
— |
| Refactor | Low |
Removes docs-staleness treemap; replaces with single doc graph SVG. Removes docs-staleness treemap; replaces with single doc graph SVG. Source: granite4.1:30b@2026-05-28-audit Confidence: low |
— |
Full changelog
/assess: one doc map, colour-blind-safe
Replaces the docs-staleness treemap with a single doc-navigability graph that carries both Layer 0 doc signals at once — and makes every /assess artifact colour-blind-safe by default.
The doc graph uses one visual channel per signal:
- Structure = navigability — radial rings by link-distance from the entry point (centre = entry, rim = unreachable). Orphans get a dashed ring, the entry a blue ring.
- Colour = staleness — OrRd (no red-green): vivid red = a frozen doc beside churning code = a lying map.
- Size = file length.
- Ghost nodes — broken links (targets that don't exist) are drawn as labelled ghosts; the missing name is the suggested fix.
It adapts the "Lint" health-check from Karpathy's LLM-wiki pattern — hubs, orphans, connectivity, dangling references, missing cross-references — made deterministic. Contradictions between pages stay an LLM/Phase-2 job, and the report says so.
Highlights
- Single doc artifact (
doc-graph.svg); docs-staleness treemap removed. - Colour-blind-safe by default: OrRd ramp for the doc graph and the code complexity heatmap; Okabe-Ito + ring/dash cues.
- Clickable, scalable SVG embeds in the report (open on their own for hover).
- Scans only repo files — git-tracked + a symlink-escape guard, so an untracked personal
CLAUDE.mdor a symlink pointing outside the repo is ignored. - Report prose explains stale/churn/centrality in plain language and distrusts the repo-baseline ranking (no more fresh-hub false positives).
- README and scoring swept to the
/8model.
Both the Claude Code plugin and the standalone Desktop/Cowork ZIPs are updated (ZIPs on the standalone-skills-latest rolling release).
What's Changed
Features
- feat: Unified doc-navigability graph for /assess (colour-blind-safe, /8 sweep) by @bjcoombs in https://github.com/bjcoombs/ai-native-toolkit/pull/33
Full Changelog: https://github.com/bjcoombs/ai-native-toolkit/compare/v1.8.0...v1.9.0
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About AI readiness toolkit
All releases →Related context
Related tools
Beta — feedback welcome: [email protected]