thumbgate 1.27.17 → 1.27.19

package/.claude-plugin/marketplace.json

@@ -0,0 +1,85 @@
+{
+  "name": "thumbgate-marketplace",
+  "version": "1.27.15",
+  "owner": {
+    "name": "Igor Ganapolsky",
+    "email": "ig5973700@gmail.com"
+  },
+  "plugins": [
+    {
+      "name": "thumbgate",
+      "description": "One 👎 becomes a hard rule the agent cannot bypass. Captures thumbs-down feedback, distills it into PreToolUse Pre-Action Checks, enforced across every future Claude Code session.",
+      "longDescription": "ThumbGate is a tripwire, not a memory layer. When you thumbs-down an agent action — a force-push that overwrote a teammate's commit, a destructive SQL drop, a refactor you already rejected — ThumbGate captures the pattern, distills a one-sentence lesson, and installs it as a PreToolUse block. The next time the agent reaches for the same tool call shape, the hook fires before execution and Claude sees your own words back as the reason.\n\nThe enforcement lives in the hook, not in the model's context, so the agent cannot bypass it by forgetting, by being prompted around it, or by switching sessions. Same rule works in Claude Code, Cursor, Codex, Gemini CLI, Amp, Cline, and OpenCode.\n\n33 pre-action checks shipped (force-push, destructive SQL, mass-delete, npm publish from wrong branch, deploy without verification). Budget enforcement (action count + time limits) prevents runaway agent sessions. NIST / SOC2 / OWASP / CWE compliance tags on every rule for enterprise teams. DPO training-pair export for teams running their own fine-tunes.\n\nFree tier: 2 feedback captures/day, 10 total captures, 3 active auto-promoted prevention rules, fully offline. Pro ($19/mo): hosted sync across machines and agent runtimes, dashboard, audit export.",
+      "source": {
+        "source": "npm",
+        "package": "thumbgate"
+      },
+      "version": "1.27.15",
+      "author": {
+        "name": "Igor Ganapolsky",
+        "email": "ig5973700@gmail.com",
+        "url": "https://thumbgate.ai"
+      },
+      "homepage": "https://thumbgate.ai",
+      "repository": "https://github.com/IgorGanapolsky/ThumbGate",
+      "license": "MIT",
+      "category": "developer-tools",
+      "tags": [
+        "guardrails",
+        "pretooluse",
+        "hooks",
+        "feedback",
+        "rlhf",
+        "dpo",
+        "agent-safety",
+        "workflow-hardening"
+      ],
+      "keywords": [
+        "guardrails",
+        "pretooluse",
+        "hooks",
+        "feedback",
+        "rlhf",
+        "dpo",
+        "agent-safety",
+        "workflow-hardening",
+        "claude-desktop",
+        "desktop-extension",
+        "pre-action-checks",
+        "ai-agent-safety",
+        "mcp",
+        "memory"
+      ],
+      "capabilities": {
+        "skills": 2,
+        "commands": 5,
+        "agents": 1,
+        "hooks": 3,
+        "mcpServer": "thumbgate serve"
+      },
+      "installCommand": "/plugin install thumbgate@claude-plugins-official",
+      "metadata": {
+        "author": "Igor Ganapolsky",
+        "homepage": "https://thumbgate.ai",
+        "license": "MIT",
+        "keywords": [
+          "guardrails",
+          "pretooluse",
+          "hooks",
+          "feedback",
+          "rlhf",
+          "dpo",
+          "agent-safety",
+          "workflow-hardening",
+          "claude-desktop",
+          "desktop-extension",
+          "pre-action-checks",
+          "ai-agent-safety",
+          "mcp",
+          "memory"
+        ],
+        "category": "developer-tools"
+      }
+    }
+  ]
+}

package/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "thumbgate",
   "description": "One 👎 becomes a hard rule the agent cannot bypass. Captures thumbs-down feedback, distills it into PreToolUse Pre-Action Checks, enforced across every future Claude Code session.",
-  "version": "1.27.17",
+  "version": "1.27.15",
   "author": {
     "name": "Igor Ganapolsky",
     "email": "ig5973700@gmail.com",

package/.well-known/mcp/server-card.json

@@ -1,6 +1,6 @@
 {
   "name": "thumbgate",
-  "version": "1.27.17",
+  "version": "1.27.15",
   "description": "ThumbGate — 👍👎 feedback that teaches your AI agent. Thumbs down a mistake, it never happens again.",
   "homepage": "https://thumbgate.ai",
   "transport": "stdio",

package/README.md

@@ -6,11 +6,8 @@
   </a>
 </p>
 
-<p align="center">
-  <a href="https://buymeacoffee.com/igorganapolsky">
-    <img src="https://cdn.buymeacoffee.com/buttons/v2/default-yellow.png" alt="Buy Me A Coffee" height="40" />
-  </a>
-</p>
+> **⚡ LAUNCH PROMO: 50% off Pro for 3 months — use code `THUMB50`**
+> [$9.50/mo instead of $19 →](https://buy.stripe.com/eVqeVd62o0aO19v7mb3sI3g) · Limited to first 50 customers
 
 **AI coding agents repeat mistakes — and one wrong tool call can wipe a directory, leak a key, or push broken code.**
 
@@ -649,6 +646,37 @@ Pro ($19/mo or $149/yr) removes the rule cap and adds history-aware lesson recal
 
 ---
 
+---
+
+## ThumbGate Pro — for Teams
+
+ThumbGate is free and MIT-licensed forever. For teams that need more:
+
+- **Team-wide enforcement policies** — apply rules across all agents and developers
+- **Centralized feedback memory** — share prevention rules across your org
+- **Budget monitoring** — track and cap agent spend per project
+- **Priority support** — direct access, SLA, onboarding help
+
+**$19/month · [Get started →](https://buy.stripe.com/4gM5kD9eA7DgdWh21R3sI3d)**
+
+---
+
+## Who builds ThumbGate — and hiring me
+
+I'm **Igor Ganapolsky** — I designed and maintain ThumbGate. If you're shipping **payments, AI agents, or Android features** and want them built by someone demonstrably careful with production and with money, I take a small number of **freelance / contract** engagements.
+
+ThumbGate is the receipt, not the pitch: it fails *closed* on dangerous agent actions and publishes a [threat model](THREAT_MODEL.md) stating exactly what it enforces and what it can't contain. Documenting where my own guardrails end is the standard I hold client work to.
+
+- **Payments** — Stripe / Stripe Connect: destination charges, split payouts, escrow & milestone release, 3DS/SCA, idempotent webhooks, reconciliation.
+- **Applied AI / agents** — tool-use guardrails, MCP servers, orchestration, and evaluation loops (the engineering behind this repo).
+- **Android + backend** — native Android and the APIs behind it, shipped end-to-end.
+
+**$120–150/hr, 1099 · remote, US timezones** → **[LinkedIn](https://www.linkedin.com/in/igor-ganapolsky-859317343/)** · **[thumbgate.ai](https://thumbgate.ai)**
+
+> ThumbGate is free and MIT-licensed, and stays that way. If it saved you a costly mistake, the best thank-you is an intro to someone who needs an engineer who ships carefully.
+
+---
+
 ## License
 
 MIT. See [LICENSE](LICENSE).