Changelog

What's new

Platform updates, new skills, and version history. Follow along as we ship.

1306+Skills

19PowerPacks

5Domains

30Updates

Platform Updates

February 19, 2026Security

Skill Delivery Guard

3-layer integrity check on every skill delivery: SHA-256 hash verification, keyword blocklist with context-aware safe patterns, and prompt injection detection. Blocks tampered or compromised content before it reaches your agent.

February 19, 2026Feature

Remote MCP Server

MCP endpoint upgraded to Streamable HTTP transport. Connect from Claude Code, ChatGPT, or any MCP client with a single URL. No local server needed. SSE keepalive, Bearer token auth.

February 19, 2026Feature

OpenClaw SKILL.md Export

New endpoint: GET /api/v1/skills/{slug}/openclaw returns a ready-to-save SKILL.md file with YAML frontmatter. Drop it into ~/.openclaw/skills/ and go. Also: ?format=text on the prompt endpoint for plain-text delivery.

February 19, 2026Content

Pipeline v4 Pilot

Next-generation skill pipeline with improved prompt engineering and research depth. Pilot batch shows +2.17 average score improvement over v3 skills. Rolling out to new skills.

February 19, 2026Security

135 Tests, 36 Security Checks

Test suite expanded to 135 tests across 14 categories. 36 dedicated security checks covering API auth, rate limiting, input validation, and content integrity. Runs on every deploy via GitHub Actions.

February 19, 2026Feature

"Built for Trust" on API Page

New section on /features/api documenting the security model: integrity verification, access control, key hashing, and continuous testing. Transparency about how skills are protected.

February 16, 2026Feature

Slot Activation System

Skills are now activated into time-based slots (7 days) instead of per-load counting. Activate, use unlimited, deactivate anytime. Free (3 slots), Pro (42), Max (100).

February 16, 2026Content

16 PowerPacks Published

Curated skill bundles for complex workflows, from Indie Hacker Starter to Enterprise Architecture Review. Each pack combines 3-6 expert skills designed to work together.

February 15, 2026Compliance

Source Attribution & IP Audit

Every skill now displays research sources with type badges, author, and year. Automated IP screening via Claude Sonnet ensures no copyright or trademark issues.

February 15, 2026Feature

Hybrid Search

Semantic search powered by pgvector (1536-dim embeddings) combined with full-text search. Weighted hybrid scoring (FTS 0.3 + cosine 0.7) with automatic fallback chain.

February 14, 2026Feature

MCP Protocol: Native Claude Code Integration

JSON-RPC 2.0 endpoint with SSE streaming. 5 tools: search_skills, load_skill, load_powerpack, list_categories, suggest_skill. Configure supaskills as an MCP server directly in Claude Code.

February 14, 2026Feature

SupaScore Quality System

Every skill scored across 6 dimensions: Research Quality, Prompt Engineering, Practical Utility, Completeness, User Satisfaction, Decision Usefulness. Tiers from Bronze to Diamond.

February 13, 2026Feature

OAuth: GitHub & Google Sign-In

One-click authentication via GitHub or Google. Same-email accounts auto-merge. No password required.

February 12, 2026Feature

REST API v1

Full API with Bearer token auth (sk_supa_ prefix). Endpoints for skill search, activation, feedback, categories, and PowerPacks. Rate-limited per plan.

February 11, 2026Launch

Platform Launch

supaskills goes live with expert skills across 6 domains and 17 categories. Every skill research-backed, quality-scored, and ready to activate via API or MCP.