Data basis: GitHub Trending bot issues — today huiyishalou/ai-trending-hub #211 (2026-11-05 16:16 UTC; star values are unlabeled but match stars-today in prior cross-checks) vs. the same bot’s #210 (2026-11-04 16:01 UTC), with kouweizhu/agents-radar #810 as a cross-reference. Primary metric is stars today; repos with 40k+ total stars, long-running chart leaders, slowing repos and non-AI repos are excluded. Descriptions come from each README; numbers come from the bot issues.
NVIDIA/auto-ontology — +4,945 ⭐ today (2 days in a row)
URL: https://github.com/NVIDIA/auto-ontology
Total: not listed by the bot (about 4.4k on the GitHub page a day earlier, plus growth since) · Day-over-day: +3,767 → +4,945 (+31%, same bot) · Python backend + Next.js · Apache-2.0
An open context layer that helps AI agents stop producing wrong numbers over enterprise data. It runs as a service beside your database, compiles DB metadata into an ontology (semantic layer), and makes agents query through it, generating SQL from what a field means rather than what a column is called. Key components:
- Auto-drafts an ontology from DB metadata, maps business terms to columns and infers table relationships
- Stores reviewed SQL for measures so metrics are computed consistently; definitions can be reviewed, edited and certified in a web UI/API
- Checks whether a question is answerable from the data before querying, and returns SQL, steps and result rows with each answer
- Next.js frontend, FastAPI (LangChain) backend, read-only MCP server on FastMCP, Postgres + pgvector
- Ten sources including PostgreSQL, MySQL, Databricks, Snowflake, ClickHouse, Trino and DuckDB; defaults to NVIDIA NIM (Nemotron) but works with OpenAI-compatible endpoints
Its second-day gain beat the first, a sign of strong demand for metric-definition governance in agentic analytics.
Practical takeaway: Worth a PoC as the governance layer that enforces approved metric definitions when you connect agents like Claude or Cursor to your data warehouse.
#SemanticLayer #TextToSQL #KnowledgeGraph #DataAgents #MCPServer
scenario-labs/skills — +1,258 ⭐ today
URL: https://github.com/scenario-labs/skills
Total: not listed by the bot (about 1.7k on a possibly stale GitHub page) · Day-over-day: +448 → +1,258 (+181%, same bot) · Python/Node.js · MIT
A collection of 100+ agent skills that teach coding agents such as Claude Code, Cursor and Codex to produce images, video, audio, textures, skyboxes and 3D assets. It works two ways: Scenario skills connect to the Scenario MCP server (OAuth or API key) and follow a discover → schema → run → wait → display → download loop, while Expert tools are lead-plus-specialist skill teams that drive locally installed creative tools and game engines without MCP. Main categories:
- Images: text-to-image, editing, product shots, brand kits, storyboards, thumbnails
- Game art and 3D: sprites, textures, skyboxes, 3D meshes, animation
- Video and audio: generation and editing, color grading, music, sound effects, voice, captions
- Consistency and custom models: character/style consistency, custom model training, per-model cost estimates and selection
- Expert tools: skill teams for Blender, Maya, ZBrush, Unreal, Unity, Godot and Roblox Studio
- Installs via
npx skills add, with Claude and Codex plugin folders included
As the agent-skills ecosystem spreads beyond coding into creative and game production pipelines, its daily gain nearly tripled.
Practical takeaway: A good starting point for marketing or game teams testing a single-agent pipeline from concept to asset to engine import (generation requires a paid Scenario account).
#AgentSkills #GenerativeMedia #GameAssets #3DGeneration #MCPServer
max-sixty/worktrunk — +1,004 ⭐ today (2 days in a row)
URL: https://github.com/max-sixty/worktrunk
Total: 10,377+ (Rain1er, Nov 3) · Day-over-day: +964 → +1,004 (+4%, same bot) · Rust · MIT OR Apache-2.0
A Git worktree management CLI for developers running several AI coding agents at once. Each task gets its own worktree so parallel agents never overwrite each other’s changes; worktrees are addressed by branch name and their paths are generated from templates. Core commands and features:
wt switch -c <branch> -x claude -- 'task': create a worktree and launch an agent in one stepwt list --full: changes, commits ahead of main, CI status and AI-generated summarieswt merge: squash, rebase, merge and clean up in one go;wt removedeletes worktree and branch- Hooks, diff-based LLM commit messages,
wt switch pr:123PR checkout - Shared build caches (
target/,node_modules/) and a unique dev-server port per worktree
As parallel-agent workflows spread, it has held around +1,000 stars a day for two straight days.
Practical takeaway: Teams running agents per feature in one repo can adopt it immediately to standardize branch, worktree and port management.
#GitWorktree #ParallelAgents #ClaudeCode #DeveloperTools #RustLang
Excluded (summary): vercel-labs/agent-desktop slowed from +4,490 to +3,968 (-12%, same bot; agents-radar +3,791 → +2,470) and falls outside the ±10% rule. mattpocock/skills (311k), D4Vinci/Scrapling (94k), claude-mem (112k) and MoneyPrinterTurbo (139k) exceed 40k total stars. vercel-labs/agent-browser (+383), ConardLi/garden-skills (+356) and pr-agent (+280) gained too little, kimi-k3-in-c (+351) is slowing, and floci, spec-kit and 90DaysOfDevOps are non-AI or perennial large repos.
Leave a Reply