chore(plugin): retire .claude-plugin/ references (GRA-1249 / GH #206)

Gradata/.dockerignore

@@ -2,7 +2,6 @@
 .git
 .github
 .claude
-.claude-plugin
 .tmp
 .venv
 venv

Gradata/CHANGELOG.md

@@ -2,6 +2,13 @@
 
 ## Unreleased
 
+### Breaking
+
+- **Plugin marketplace install path retired.** The `.claude-plugin/` directory
+  and `/plugin install gradata` flow are gone; use
+  `gradata install --agent claude-code` instead. See PR #211 for the README
+  guide. (GRA-1249, epic GRA-1198)
+
 ### Changed
 
 - Renamed the deterministic anchor-preserving prompt injection module to

Gradata/examples/README.md

@@ -33,4 +33,5 @@ pip install gradata
 python examples/with_claude_code.py
 ```
 
-For the zero-code install, see [.claude-plugin/README.md](../.claude-plugin/README.md).
+For the zero-code install, run `gradata install --agent claude-code` — see the
+top-level [README](../README.md#install-for-claude-code) for details.

Gradata/examples/with_claude_code.py

@@ -1,17 +1,17 @@
 """Using Gradata alongside Claude Code.
 
-With the Claude Code plugin installed (`/plugin install gradata`),
-corrections on `Edit` / `Write` tool calls are captured automatically and
-graduated rules are injected into every session. No code is needed in the
-hot path.
+With `gradata install --agent claude-code` run once, corrections on `Edit` /
+`Write` tool calls are captured automatically and graduated rules are
+injected into every session. No code is needed in the hot path.
 
 This example is the manual equivalent: the same `Brain` API the hooks use,
 run directly so you can inspect what gets stored and what would be injected.
 
 Run:
     python examples/with_claude_code.py
 
-See also: .claude-plugin/README.md for the zero-code install flow.
+See also: the README's "Install for Claude Code" section for the zero-code
+install flow.
 """
 
 from pathlib import Path