Personal monorepo and Claude Code workspace for Henry Goldsmith. Contains multiple production-ready web apps, automation scripts, a self-improving knowledge base, and extensive Claude agent skills/commands.
| Path | Purpose |
|---|---|
apps/ |
Standalone Next.js / Vite / Python / Electron applications |
.claude/ |
Claude Code skills, agents, commands, helpers, settings |
src/workspace_daemon/ |
Background workspace daemon (Google/MSFT integration) |
execution/ |
One-off Python automation scripts (leads, email, invoices, etc.) |
raw/ |
Append-only knowledge ingestion zone |
wiki/ |
Indexed knowledge base (read this first) |
output/ |
Improvement-loop artifacts and review files |
CLAUDE.md |
Project rules, agent coordination, swarm config |
Source of truth: apps/registry.json (machine-readable, validated by scripts/validate-registry.mjs). Deployment source of truth: docs/deployment-dependency-map.md, enforced by each app's Vercel ignoreCommand. Lifecycle: active = shipped, incubating = in dev, maintenance = bugfix-only, archived = frozen reference, superseded = replaced, tooling/service = internal, external = standalone repo.
| App | Path | Stack | What it is |
|---|---|---|---|
| Arise | apps/arise |
Vite + React + Tailwind | Training & progression (programs, sessions, levelled attributes) |
| Daily Debate | apps/daily-debate |
Next.js + Supabase + Anthropic | Daily critical-thinking debates (solo vs AI + PvP) with argument graph |
| Reflect | apps/emotion-tracker · site apps/reflect-site |
Next.js + Anthropic | Structured reflection event→assumptions→alternatives→action→follow-up with hedged bias language + longitudinal calibration |
| Forq | apps/food-shopping-os · site apps/forq-site |
Next.js (local-first) | Food shopping & planning OS (pantry, retailer, nutrition) |
| Le Studio French | apps/french-practice · hub apps/le-studio-site |
Vite + React + Tailwind | French PWA (Today/Speak/Review/Learn/Progress, Groq arena) — family hub is Le Studio/Forq/Chrono |
| Habit | apps/habit-tracker |
Next.js + Supabase | Quiet habit tracker with daily check-ins, targets and Pulse history |
| Revise | apps/revise |
Next.js + Supabase + FSRS | GCSE + A-level revision across WJEC/AQA/Edexcel/OCR (2,216 specPoints, validator) |
| Rapport | apps/rapport |
Next.js (local-first) + optional Supabase | Adaptive social-skills training — skill graph, mastery model, conversation simulator, real-world challenges (works with no AI provider) |
| Noticed | apps/mental-load-tracker |
Next.js + Supabase Realtime | Shared board for household invisible labor (two-person, no leaderboard) |
| rtk | apps/rtk |
Node CLI | Token-saving filter for noisy tool output (per-tool parsers + redaction) |
Also in this repo (not product apps, but tracked so the registry doesn't drift): packages/le-studio-tokens (shared design tokens), src/workspace_daemon (background sync service), .claude (skills/agents/helpers incl. prompt-improver), vendor/life-os-scrape (archived — frozen Life OS build, superseded by arise; see its README). External: Chrono (henrygoldsmith07-wq/chrono-calendar, ICS/JSON interop).
| Document | Purpose |
|---|---|
docs/definition-of-done.md |
Minimum quality gates per product (install, lint, type-check, tests, build, E2E, security, perf) |
docs/ci-coverage-audit.md |
Which workflow runs which gate for which app; live gap list |
docs/ai-standard.md |
Shared AI pipeline: validation → context → provider → schema → deterministic gates → confidence → provenance → output |
docs/security-standard.md |
Auth/RLS/tenancy/secrets/SSRF/rate-limit audit + enforcement |
docs/observability.md · packages/observability/ |
Privacy-safe AI telemetry & error records (no user content by construction) |
docs/data-provenance.md |
Source/timestamp/freshness/transformation/confidence/version for derived insights |
docs/migration-backup.md |
Migration re-runnability, RLS re-verification, backup/export/recovery per DB-backed app |
docs/a11y-baseline.md · docs/a11y.md |
Accessibility baseline + Le Studio family checklist |
docs/perf-budgets.md · scripts/performance-budgets.json |
Bundle budgets enforced in CI |
docs/evidence-registry.md · evidence/registry.json |
Durable claim ledger: demonstrated → externally validated, never auto-upgraded |
docs/validation-state.md |
Per product: implemented vs proven effective — kept strictly separate |
See CLAUDE.md for the full contract. Key commands:
/data-ingestion— Sync sessions, ecosystem data, curated content →raw/+wiki//improve-system— Auto-apply low-risk hygiene; propose higher-stakes changes for sign-off
Routines run Tue/Fri (UK time) on the knowledge-base-updates branch.
claude mcp add claude-flow -- npx -y ruflo@latest mcp start
npx ruflo@latest doctor --fixMIT — see LICENSE.