Session logs and historical notes. Active work items are tracked in GitHub Projects.
- BL-055: Fix
/committo branch before committing (PR #124) - Added AI context warning to
/commitStep 7 - Added
requires:human-reviewlabel auto-detection to/backlog add - Added
save-decisions-earlyrule - Created BL-056: Context hygiene lint script + weekly audit
- Created BL-057: Sync agent-skill patterns to starter kit
- PR #127: human-review detection — waiting on CI
- Fixed 4 stale backlog items stuck as "In Progress" on project board (BL-030, BL-035, BL-036, BL-037)
- Added project board sync step to
/commitskill (PR #118) - Built
/parallelskill for multi-agent concurrent work (PR #119) - Created BL-052–BL-055: agent/skill redesign + commit fix backlog items
- Supabase migrations pushed (001-008)
-
/supabase-syncskill created - Skills format fix — all 17 skills converted to directory format
- Rules vs References split — moved 4 rarely-used rules to references
- Nix + direnv setup
- CLAUDE.md slimmed from 308 to 106 lines
- Tightened RLS policies for
bubble_scores - Fixed timer cleanup leak in WordBubble
- WelcomeGate redirect fix
- Browser locale detection
- Extracted
useBubbleGamehook
- Bubble session redesign with multi-nickname DB model
- Guest vs signed-in scoring
- Clickable leaderboard rows
- Migration 006-008
- Floating Danish word bubbles with click-to-discover
- Game leaderboard with localStorage + Supabase sync
- Trophy badge + GamePanel inline
- Support redesign (renamed from "Donate")
- Dictionary inflections from DDO
- In-App Feedback with Supabase + mailto fallback
- WhatsNew banner, Drill mode, Vocabulary pagination
- Study mode Daily/All toggle
- Vocabulary inflection tables
- Exercises: 102 to 201; Words: 71 to 277
- Scraping + enrichment scripts
- README, Vercel Analytics, lazy-loaded routes
- SpeakSpeak + Gyldendal scrapers
- Initial 35 exercises scraped from PD3M2
- 143 verbs with empty inflections — run
enrich-vocabulary.pywithANTHROPIC_API_KEY - WordOrder UX — no reorder without clearing; needs click-to-remove or dnd
- Writing & Speaking modes — need
ANTHROPIC_API_KEYfor AI scoring