Add nuhuh plugin under Code Quality Testing

README.md

@@ -91,6 +91,7 @@ Install or disable them dynamically with the `/plugin` command — enabling you
 - [debug-session](./plugins/debug-session)
 - [debugger](./plugins/debugger)
 - [double-check](./plugins/double-check)
+- [nuhuh](./plugins/nuhuh)
 - [optimize](./plugins/optimize)
 - [performance-benchmarker](./plugins/performance-benchmarker)
 - [refractor](./plugins/refractor)

plugins/nuhuh/.claude-plugin/plugin.json

@@ -0,0 +1,9 @@
+{
+  "name": "nuhuh",
+  "description": "Your agent said Done. nuhuh runs the experiment and bounces the agent until the receipt is green.",
+  "version": "0.1.10",
+  "author": {
+    "name": "sjh9714"
+  },
+  "homepage": "https://github.com/sjh9714/nuhuh"
+}

plugins/nuhuh/README.md

@@ -0,0 +1,15 @@
+# nuhuh
+
+Your agent said "Done." nuhuh runs the experiment.
+
+This plugin installs a Stop hook. Every time the agent tries to finish, nuhuh extracts the completion claims from its final message ("all tests pass", "created src/x.ts", "endpoint works") and re-runs reality, fresh. The whole test suite in a clean process, the build, the file on disk, the local endpoint. If a claim is false, the Done is rejected and the failing evidence goes straight back to the agent, which returns to work. After 3 bounces it hands the receipt to you.
+
+Deterministic and local. No LLM calls in the verification path, no API key, nothing leaves your machine. Claims it cannot safely check are marked unverifiable, never failed.
+
+Full documentation, a False Done Rate benchmark, and the source live at [github.com/sjh9714/nuhuh](https://github.com/sjh9714/nuhuh).
+
+Try it without installing anything
+
+```bash
+npx nuhuh demo
+```

plugins/nuhuh/hooks/hooks.json

@@ -0,0 +1,15 @@
+{
+  "hooks": {
+    "Stop": [
+      {
+        "hooks": [
+          {
+            "type": "command",
+            "command": "npx --yes nuhuh@0.1.10 gate",
+            "timeout": 600
+          }
+        ]
+      }
+    ]
+  }
+}

plugins/nuhuh/plugin.json

@@ -0,0 +1,9 @@
+{
+  "name": "nuhuh",
+  "description": "A Stop hook that verifies the agent's completion claims by re-running reality. It extracts every claim from the final message (tests pass, file created, endpoint works) and verifies each one fresh with real exit codes, then bounces a false Done back to the agent with the failing evidence.",
+  "version": "0.1.10",
+  "author": {
+    "name": "sjh9714"
+  },
+  "homepage": "https://github.com/sjh9714/nuhuh"
+}