diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index ab50004..edeaf1b 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -15,7 +15,7 @@ concurrency:
 env:
   # Pinned so a new Claude Code release cannot change what `validate --strict` accepts
   # without a reviewed commit. Bump it deliberately.
-  CLAUDE_CODE_VERSION: 2.1.311
+  CLAUDE_CODE_VERSION: 2.1.312
 
 jobs:
   test:
diff --git a/CHANGELOG.md b/CHANGELOG.md
index d629050..eba6976 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -29,7 +29,8 @@
     tokens, unchanged, because it counts the four agents and omits the workflow. `/context`
     counts the agents at 169 tokens. It counts the workflow's entry for the installed
     0.3.1 (1,303 characters) at ~440–460 tokens, so the entry costs about three times
-    what `plugin details` reports. At the same ratio, 0.5.0's entry is about 320 tokens.
+    what `plugin details` reports. With 0.5.0 installed, `/context` counts its
+    932-character entry at ~320 tokens.
 - README: a new "When to use which" section sets review-kit against the built-in
   `/code-review`. Before a capped run documents `estimate`, the token-budget section gives
   the new sizing, and Cost notes add measured figures:
diff --git a/README.md b/README.md
index 89989b0..97d6933 100644
--- a/README.md
+++ b/README.md
@@ -114,7 +114,7 @@ branch or PR.
 ### Requirements
 
 - Claude Code 2.1.305 or later, with [dynamic workflows](https://code.claude.com/docs/en/workflows)
-  turned on (on Pro, the Dynamic workflows row in `/config`). Tested on 2.1.311. 2.1.305
+  turned on (on Pro, the Dynamic workflows row in `/config`). Tested on 2.1.312. 2.1.305
   is the first release that resumes a workflow whose pipeline items run more than one
   agent reliably, which [continuing a run](#continuing-a-run) relies on. The agents also
   use `omitClaudeMd` (v2.1.271+) and a per-agent cache TTL (v2.1.248+).