feat(plugin): a Claude Code plugin for context-guru, with the #141 review's blockers fixed

.claude-plugin/marketplace.json

@@ -0,0 +1,21 @@
+{
+  "name": "context-guru",
+  "description": "Run context-guru in front of your Claude Code sessions: install once, route per repo.",
+  "owner": {
+    "name": "rossoctl",
+    "url": "https://github.com/rossoctl/context-guru"
+  },
+  "plugins": [
+    {
+      "name": "context-guru",
+      "source": "./context-guru-plugin",
+      "displayName": "context-guru",
+      "description": "Install, route, inspect and remove a local context-guru proxy for Claude Code. Recovers prompt-cache misses on long sessions; no API key needed on a Pro/Max subscription.",
+      "homepage": "https://github.com/rossoctl/context-guru",
+      "repository": "https://github.com/rossoctl/context-guru",
+      "license": "Apache-2.0",
+      "keywords": ["cache", "cost", "proxy", "tokens", "prompt-caching"],
+      "category": "productivity"
+    }
+  ]
+}

README.md

@@ -108,7 +108,25 @@ docker build -t context-guru:local .
 
 ## Quickstart (60 seconds)
 
-Download a release binary — statically linked, **no Go and no C compiler needed** — or build
+**Claude Code users — install once per machine, route once per repo; no toolchain, and no API key
+needed on a Pro/Max subscription** ([details](docs/how-to/install-plugin.md)):
+
+```
+/plugin marketplace add rossoctl/context-guru
+/plugin install context-guru@context-guru
+/reload-plugins
+/context-guru:install
+```
+
+`/reload-plugins` is what makes the `/context-guru:*` skills exist in this session; without it the
+last line answers `Unknown command`. A new session does the same thing.
+
+That installs a statically-linked binary (no Go, no C compiler), routes **this project only** by
+default, starts the proxy on demand and lets it exit when idle. `/context-guru:uninstall` undoes it,
+restoring any base URL it replaced. The plugin installs with `--preset cache` — the prompt-cache
+split and nothing else. (The proxy's own default is `house`; `--preset` is how you change it.)
+
+Or by hand — a release binary is statically linked, **no Go and no C compiler needed** — or build
 from source:
 
 ```sh

context-guru-plugin/.claude-plugin/plugin.json

@@ -0,0 +1,48 @@
+{
+  "name": "context-guru",
+  "displayName": "context-guru",
+  "description": "Installs a local context-guru proxy, routes this project's Claude Code sessions through it, and reports what it saved. Reversible: /context-guru:uninstall removes the one settings key it adds and stops the proxy.",
+  "version": "0.1.0",
+  "author": {
+    "name": "rossoctl",
+    "url": "https://github.com/rossoctl/context-guru"
+  },
+  "homepage": "https://github.com/rossoctl/context-guru",
+  "repository": "https://github.com/rossoctl/context-guru",
+  "license": "Apache-2.0",
+  "keywords": [
+    "cache",
+    "cost",
+    "proxy",
+    "tokens",
+    "prompt-caching"
+  ],
+  "userConfig": {
+    "port": {
+      "type": "number",
+      "title": "Proxy port",
+      "description": "Local port the proxy listens on. The port must be FIXED rather than negotiated: the ANTHROPIC_BASE_URL written into your settings and the session hook that starts the proxy have to agree, and a URL cannot be renegotiated after it is written. Default 8787 \u2014 deliberately not 4000, which collides with litellm.",
+      "default": 8787,
+      "min": 1024,
+      "max": 65535
+    },
+    "preset": {
+      "type": "string",
+      "title": "Preset",
+      "description": "Compaction pipeline. `cache` (the default) runs the prompt-cache split and NOTHING else: no content dropped, no markers, no extra tool, no model calls. `codesmart` adds the offloaders once you want them.",
+      "default": "cache"
+    },
+    "idle_exit": {
+      "type": "string",
+      "title": "Idle exit",
+      "description": "Exit the proxy after this long with no requests and no keep-alive ping pending, so nothing is left running on your machine. Must be at least 2x the store's entry lifetime (~5h34m at the default), because exiting clears in-memory cache state.",
+      "default": "24h"
+    },
+    "upstream": {
+      "title": "Upstream base URL",
+      "type": "string",
+      "default": "",
+      "description": "Forward to this Anthropic-compatible base URL instead of api.anthropic.com. Set it when something else is already the gateway \u2014 a hosted agent pod, a corporate gateway \u2014 so the proxy CHAINS behind it rather than replacing it. That gateway keeps holding the credential and, on at least one pod, rewriting model names: bypassing it sends model ids the real API has never heard of and every request fails. Empty means go straight to Anthropic."
+    }
+  }
+}

context-guru-plugin/hooks/hooks.json

@@ -0,0 +1,26 @@
+{
+  "hooks": {
+    "UserPromptSubmit": [
+      {
+        "hooks": [
+          {
+            "type": "command",
+            "command": "\"${CLAUDE_PLUGIN_ROOT}\"/scripts/check-proxy.sh",
+            "timeout": 30
+          }
+        ]
+      }
+    ],
+    "SessionStart": [
+      {
+        "hooks": [
+          {
+            "type": "command",
+            "command": "\"${CLAUDE_PLUGIN_ROOT}\"/scripts/start-proxy.sh",
+            "timeout": 60
+          }
+        ]
+      }
+    ]
+  }
+}