agentrigs
+ submit rig
~ / newsletter / #1
// The Rig Report · 2026-10-09

The Rig Report #1

This week in public AI-agent setups: 2,169 of the 6,413 rigs we index were updated in the last 7 days. Here's what stood out.

Stat of the week

73%
of 6,413 public AI-agent rigs score 0/5 on guardrails: no deny rules, no pre-tool hook, nothing. Only 34 score 5/5. More →

Fresh rigs worth stealing from

Updated this week, well guarded (3+/5) and lean.

3/5 · 4.7k tokgastownhall/beads ★ 27.7k · Pragmatist
Beads - A memory upgrade for your coding agent
4/5 · 11.6k tokDonchitos/Claude-Code-Game-Studios ★ 25.9k · Fort Knox
Turn Claude Code into a full game dev studio — 49 AI agents, 72 workflow skills, and a complete coordination system mirroring real studio hierarchy.
3/5 · 6.4k tokmksglu/context-mode ★ 25.5k · Pragmatist
Context window optimization for AI coding agents. Sandboxes tool output (98% reduction), persists session memory, and enforces routing across 17 platforms via
3/5 · 10.1k tokpromptfoo/promptfoo ★ 25.8k · Skill Collector
Test your prompts, agents, and RAGs. Red teaming/pentesting/vulnerability scanning for AI. Compare performance of GPT, Claude, Gemini, DeepSeek, and more. Simpl
5/5 · 4.2k tokvinta/hal-9000 ★ 138 · Fort Knox
Opinionated AI coding agent and dev environment automation for macOS

Rising tools

  • mcp.amplitude.com (MCP) in 6 rigs from 6 owners · 0.17% of active rigs vs 0% of older ones (×5.5)
  • mcp.grep.app (MCP) in 11 rigs from 10 owners · 0.28% of active rigs vs 0.04% of older ones (×4.3)
  • mcp.hubspot.com (MCP) in 5 rigs from 5 owners · 0.14% of active rigs vs 0% of older ones (×4.7)
  • typesafe-ai (skill) in 17 rigs from 17 owners · 0.47% of active rigs vs 0% of older ones (×14.2)
  • diagnosing-bugs (skill) in 53 rigs from 53 owners · 1.39% of active rigs vs 0.11% of older ones (×10.1)
  • test-audit (skill) in 16 rigs from 16 owners · 0.45% of active rigs vs 0% of older ones (×13.4)

Full trending list and method →

Guardrail tip: Deny secret reads before anything else

Most rigs that leak keys never meant to: the agent just ran `cat .env` while debugging. Two lines in .claude/settings.json stop it:

{
  "permissions": {
    "deny": ["Read(./.env)", "Read(./.env.*)", "Read(~/.ssh/**)", "Read(**/*.pem)"]
  }
}

Check your own rig in 30 seconds with the Rig Doctor.

Starter of the week

Frontend web: React / Next.js / Vue / Svelte UIs. AGENTS.md, guarded settings.json, guard hook and skills in one command.

archive
copied ✓