Phase 1: Core Platform — 100% COMPLETE ✓
MVP launched with all foundational features:
- Web marketplace UI
- Hybrid semantic + keyword search (vector + FTS5 + RRF fusion)
- User authentication (GitHub OAuth via Better Auth)
- Ratings & reviews system (0-10 scale)
- API key management (SHA-256 hashed)
- CLI tool (skillx) with search, use, publish, report, config commands
- Deployment on Cloudflare Workers + D1 + Vectorize stack
- Content security scanning (prompt injection, invisible chars, ANSI escapes)
- Install tracking with device ID deduplication
Phase 3.1: Leaderboard Enhancements — 100% COMPLETE ✓
- Voting system (up/down, Redis-style)
- Sort tabs (best, rating, installs, trending, newest)
- Category + risk filter dropdown
- Skill preview modal (description, category, stats)
- Scoring algorithm with 7 signals (rating 30%, installs 20%, votes 10%, rating_count 15%, recency 15%, reviews 5%, verified 5%)
- Client-side vote/favorite state overlay on KV-cached leaderboard
- Rate limiting (10 votes/min per user)
Phase 3.2: Skill References & Scripts — 100% COMPLETE ✓
- Database:
skill_referencestable (title, filename, url, type enum, content) - Database:
scriptsJSON column on skills table - Database:
fts_contentcomputed column for extended FTS5 search - GitHub fetcher: Tree API scan for SKILL.md + metadata extraction
- Seed pipeline: Top 50 skills → 500 → full rollout (progressive)
- API: Detail endpoint returns references + scripts arrays
- UI: Skill detail page displays references + scripts sections
- CLI:
--include-refsand--include-scriptsflags forskillx use - Vectorize indexing: Skill content + reference titles (200K vectors)
Phase 3.5: Skill Publishing & Registration — 100% COMPLETE ✓
- Register API:
/api/skills/registerwith auth validation - GitHub ownership verification via repo write access check
- Multi-skill scanning (single SKILL.md or repo-wide scan)
- Content sanitization & security scanning before storage
- CLI
skillx publishcommand: auto-detect owner/repo - Usage tracking via install counts and execution reports
Status: Next Priority (after Phase 3.3+) Duration: 4-6 weeks Goals: Enable paid skills, payment processing, production deployment
- Stripe account setup + API keys
- Product setup in Stripe (paid skills)
- Payment UI (card entry, checkout modal)
- Webhook handlers (payment.created, payment.failed)
- Revenue tracking database (transactions table)
- Payout calculation (70% creator, 30% platform)
- Create GitHub OAuth app (prod environment)
- Configure redirect URIs (skillx.sh domain)
- Set client ID/secret in production wrangler config
- Test full OAuth flow on production domain
- Disable development OAuth credentials
- Create production Cloudflare account
- Set up custom domain (skillx.sh) + SSL
- Create production D1 database (skillx-db-prod)
- Create production KV namespace (kv-prod)
- Create production Vectorize index (skillx-skills-prod)
- Configure environment variables (.env.production)
- Enable Cloudflare Analytics + Observability
- Backup dev D1 database
- Run migrations on production D1
- Load seed data (30 initial skills)
- Verify data integrity
- Index skills in production Vectorize
- Load testing (1K concurrent users)
- Search performance testing (p95 <800ms)
- Payment flow testing (multiple payment methods)
- Auth flow testing (OAuth + API keys)
- Security audit (OWASP Top 10)
- Bug fixes & optimizations
- Deploy to production (wrangler publish)
- Monitor error rates, latency, uptime
- Set up alerts (error spikes, slow queries)
- Create runbook for common incidents
- Document rollback procedure
- Search latency p95 <800ms in production
- 99.5% uptime (first 30 days)
- Zero critical security issues
- Payment processing working for 10+ transactions
- GitHub OAuth SSO verified
Status: Planned Duration: 7-9 weeks Goals: Leaderboard enhancements, skill references & scripts, MCP server, sandbox execution
Implementation order: Ready-to-implement plans first (3.1, 3.2), then features with external dependencies (3.3-3.6).
Status: All 4 phases implemented and merged Implementation plan: plans/260213-1558-leaderboard-enhancements/plan.md
- Phase 1: Database migration (
votestable + skills columns) + Vote API - Phase 2: Sort/filter controls + clickable author + preview modal
- Phase 3: Favorites button + vote arrows + auth overlay
- Phase 4: Scoring algorithm updates (7 signals: rating 30%, installs 20%, votes 10%, rating_count 15%, recency 15%, reviews 5%, verified 5%)
Status: All 6 phases implemented and merged Implementation plan: plans/260213-1218-skill-references-scripts/plan.md
- Phase 1: DB schema migration (
scripts,fts_contentcolumns +skill_referencestable) - Phase 2: GitHub fetcher script (Trees API, progressive rollout)
- Phase 3: Seed pipeline + Vectorize integration
- Phase 4: API & search updates (detail endpoint returns refs/scripts)
- Phase 5: UI skill detail page (references + scripts sections)
- Phase 6: CLI updates (
--include-refs,--include-scriptsflags)
Status: Auth validation + CLI publish command implemented Implementation plan: plans/260305-skillx-publish-command/plan.md
- Phase 1: Add auth + ownership validation to Register API
- Phase 2: Create CLI
publishcommand - Phase 3: Typecheck + manual verification
Features:
- GitHub repo ownership verification (write access check)
- Automatic content scanning for security
- Multi-skill registration support (single file or repo scan)
- CLI
skillx publish owner/repowith dry-run mode
- Design MCP protocol for skill discovery
- Implement MCP server in CLI tool
-
skillx-mcp-serverpackage - Claude integration via MCP
- Skill result formatting (JSON/markdown)
Example usage:
npx skillx-mcp-server
# Server listens on stdio
# Claude makes requests: { method: 'tools/call', name: 'search', args: { query } }- Fetch radar charts from Skillmark.sh API
- Embed in skill detail page
- Cache radar charts (30 min TTL)
- Display skill verification badge
- Research: Deno Deploy, AWS Lambda, or Cloudflare Workers
- Design execution environment (timeouts, resource limits)
- Implement skill runner (fetch SKILL.md, parse commands)
- Output capture (stdout, stderr, exit code)
- Security: sandboxing, env var isolation
Phase 3.1-3.3 (Complete):
- References & scripts displayed on skill detail pages
- CLI
skillx use --include-refs --include-scriptsshows all content - Leaderboard sort/filter/preview working
- Vote system functional with rate limiting (10 votes/min per user)
- Scoring algorithm updated with vote signal (7 signals total)
- Skill publishing with GitHub ownership validation
- Multi-skill registration from GitHub repos
Phase 3.4-3.6 (Pending):
- MCP server working with Claude
- 10+ skills successfully published via
skillx publish - Sandbox execution <5s per skill
- Radar charts loading & caching
Status: Future Duration: Q3-Q4 2025 Goals: Community features, advanced search, monetization
- User profiles with bio, avatar, verified badge
- Skill review comments (replies, threading)
- User reputation system (badges, achievements)
- Discussion forum per skill
- Trending skills (by rating, usage, recency)
- Personalized recommendations (based on favorites)
- Skill collections (curated bundles)
- Advanced filters (tags, language, performance)
- Subscription tiers (free, pro, enterprise)
- Usage-based pricing (per API call)
- Team management (multi-user accounts)
- API quota system (e.g., 1K searches/month)
- Skill creator dashboard (usage stats, revenue)
- User analytics (most-used skills, success rate)
- Platform analytics (total searches, revenue, growth)
| Metric | Target | Actual | Status |
|---|---|---|---|
| Skills indexed | 30+ | 30 | ✓ |
| Feature completeness | 100% | 100% | ✓ |
| Code coverage | 70%+ | TBD | ? |
| Documentation coverage | 100% | 100% | ✓ |
| Metric | Target | Status |
|---|---|---|
| Production uptime | 99.5% | Planned |
| Payment processing | 100% | Pending |
| Search latency p95 | <800ms | On track (local: <200ms) |
| GitHub OAuth production | Working | Planned |
| Metric | Target | Status |
|---|---|---|
| Skills with references | 100% | Complete (all skills) |
| Skills with scripts | 100% | Complete (where applicable) |
| Leaderboard performance | <500ms | Complete |
| Vote system uptime | 99%+ | Complete |
| Security scanning coverage | 100% | Complete (all new skills) |
| Skills published via CLI | 10+ | In progress |
| MCP server adoption | 50+ users | Planned |
| Sandbox execution success | 95%+ | Planned |
- Stripe account approval (usually <24hrs)
- Cloudflare domain setup (DNS changes)
- GitHub OAuth app creation (instant)
- Skillmark.sh API documentation
- Deno Deploy or alternative sandbox provider
- MCP protocol finalization (Claude ecosystem)
- None currently identified
| Phase | Team Size | Estimated Hours |
|---|---|---|
| Phase 1 (Complete) | 2 devs | 160h |
| Phase 2 | 2 devs | 180h |
| Phase 3 | 2-3 devs | 240h |
| Phase 4 | TBD | TBD |
Jan 2025 Feb 2025 Mar 2025 Apr 2025 May 2025
│ │ │ │ │
├─ Phase 1 ─┤
Complete
├────── Phase 2 ──────┤
Hardening & Payments
├──────────── Phase 3 ─────────┤
MCP & Execution
├────── Phase 4 ──→
Community & Advanced
| Decision | Context | Date |
|---|---|---|
| Use Cloudflare stack | Serverless, edge compute, free tier | Jan 2025 |
| Hybrid search (vector + FTS5) | Balance semantic + keyword search | Jan 2025 |
| Better Auth + GitHub OAuth | Passwordless, minimal setup | Jan 2025 |
| Stripe for payments (Phase 2) | Market-leading, good CLI/API | TBD |
| MCP server for Claude (Phase 3) | Leverage Claude ecosystem | TBD |
Last Updated: Mar 5, 2026 Next Review: Mar 12, 2026 Current Phase: Phase 3 (Features) — Leaderboard (complete), References/Scripts (complete), Publishing (complete) Next Milestone: Phase 3.4 MCP Server Implementation