Add kalibr — autonomous model + tool routing for OpenClaw agents based on real time telemetry

skills/kalibr-ai/kalibr/SKILL.md

@@ -0,0 +1,56 @@
+---
+name: kalibr
+description: Autonomous execution path routing for OpenClaw agents. Kalibr hooks into your agent's runtime, reads live telemetry, and routes to the optimal model + tool + parameter combination — preventing failures, degradations, and cost spikes before they impact users.
+homepage: https://github.com/kalibr-ai/openclaw-kalibr
+metadata:
+  clawdbot:
+    emoji: "⚡"
+    requires:
+      env: ["KALIBR_API_KEY"]
+      bins: ["node"]
+    primaryEnv: "KALIBR_API_KEY"
+    install:
+      - id: npm
+        kind: node
+        package: "@kalibr/openclaw"
+        label: Install Kalibr OpenClaw plugin (npm)
+---
+
+# Kalibr — Autonomous Execution Path Routing for OpenClaw Agents
+
+Kalibr hooks into your OpenClaw agent's runtime, reads live telemetry, and routes to the optimal execution path (model + tool + parameters) — preventing failures, degradations, and cost spikes before they impact users.
+
+## Installation
+```bash
+openclaw plugins install @kalibr/openclaw
+openclaw config set plugins.entries.kalibr.config.apiKey "your-kalibr-api-key"
+openclaw config set plugins.entries.kalibr.config.enableRouting true
+openclaw gateway restart
+```
+
+Get your API key at: https://dashboard.kalibr.systems
+
+## Two Modes
+
+**Telemetry mode** (enableRouting: false) — instruments every LLM call with zero behavior changes. Full visibility into success rates, latency, cost, and failure patterns across providers.
+
+**Routing mode** (enableRouting: true) — everything in telemetry mode, plus autonomous rerouting to the optimal model + tool + parameter combination. Uses Thompson Sampling + Wilson scoring. Detects provider degradation via 10% canary traffic and reroutes before users are affected.
+
+## OpenClaw Hooks
+
+- before_model_resolve — execution path override (primary hook)
+- before_prompt_build — context injection
+- before_agent_start — legacy fallback
+
+## Reliability
+
+Agents running in production with Kalibr see near 100% success rates even during provider incidents and degradations.
+
+## Resources
+
+- Docs: https://kalibr.systems/docs
+- Dashboard: https://dashboard.kalibr.systems
+- GitHub: https://github.com/kalibr-ai/openclaw-kalibr
+- npm: https://www.npmjs.com/package/@kalibr/openclaw
+- Python SDK: https://github.com/kalibr-ai/kalibr-sdk-python
+- SKILL.md: https://kalibr.systems/skill.md