v4.19.0
Tagged 2026-03-17. 31 changes between v4.18.0 and v4.19.0.
Reconstructed from the git tag range on 2026-08-06 during a changelog backfill — releases v4.17.0 through v4.33.0 shipped without notes. Every item below is a real commit; none of the wording infers intent beyond what the commit itself states.
Features
- box — Add box-cloud-filesystem plugin (#368)
- catalog — Add geepers + lumera-agent-memory; fix PR review bugs (#367)
- browser-compat — Add Kobiton as first-class cloud provider, replace all stubs
Fixes
- refs — Resolve final 3 audit findings (2 stubs + 1 false positive) (#365)
- wondelai — Fill implementation.md stubs with methodology guides (#363)
- refs — Add remaining reference stubs (lokalise, documenso, speak)
- refs — Fill all remaining reference stub files
- saas-packs — Expand prose in skills to meet body-substance threshold
- api-development — Fill remaining 13 reference stub files with real examples
- misc — Fill 12 reference stubs across community/crypto/productivity plugins
- devops/saas — Fill 22 reference stubs across devops and saas-pack plugins
- api-development — Fill 12 reference stub files with real examples
- openrouter-pack — Replace boilerplate with unique content in all 30 skills
- ai-ml — Fill 7 reference stub files with real code examples
- performance — Replace generic boilerplate openings in 24 skills
- ai-ml — Replace generic boilerplate openings in 21 skills
- skill-enhancers — Build out 4 empty shells + gh-dash with real content
- security/packages — Replace generic boilerplate openings in 13 skills
- audit — Eliminate 6 false positives in content quality audit
- browser-compat — Shorten command description to ≤80 chars
- browser-compat — Address PR review comments
- browser-compat — Restructure SKILL.md sections for validator A grade
- explore — Add missing ‘more’ text to tag overflow indicators
- explore — Remove verification badges and flat card outlines
- explore — Clean up card layout and remove confusing Verified Only toggle
Performance
- browser-compat — Optimize SKILL.md to 99/100 on rubric
Refactoring
- navigating-github — Focus on setup + learning, drop redundant modes
Other
- Add content quality audit script with 7 checks for boilerplate/duplicates/stubs
- Align validators with Anthropic spec: two-tier system (standard + enterprise)
- Add PM AI Partner: 12 PM-specific agent skills + 6 workflow commands (#327)