diff --git a/CHANGELOG.md b/CHANGELOG.md
index 7a6c65b..a0ab4ff 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -4,6 +4,24 @@ All notable changes to this project are documented here.
 
 ---
 
+## [Unreleased]
+
+### Changed
+
+- **Published in the OpenAI Plugins Directory** as [Avoid AI Writing](https://chatgpt.com/plugins/plugins_6a9b77b18b8881918efa9c1255868164) (version 3.29.0, approved 2026-09-04). The bundled canonical skill now omits the frontmatter `metadata` block, which the portal rejects (#146); TERMS.md and PRIVACY.md state the plugin's scope and data handling in the terms OpenAI's plugin guidelines ask for (#147).
+- **Plugin validation now fails closed on deferred port-integrity gaps.** `agents/openai.yaml` rejects scalar policies and malformed mapping/list lines, SVG assets must have an actual `<svg>` root, and the bundled routing matrix carries a checked graph digest plus generated edge inventory so it cannot silently drift from `skill-graph.json`.
+
+- **README adds a restrained related-work block.** Links to Conor's public
+  builds, Chain of Thought, and `repo-audit` now sit after the core product and
+  usage documentation.
+- **Corpus manifest documents the register gap with two auditable seed entries.**
+  RFC 8259 provides a pre-LLM `docs` source and a 1995 W3C mailing-list message
+  provides a pre-LLM `conversational` source. The corpus README records that
+  both registers remain under-sampled and that `social` and `email` still have
+  no entries; the additions do not authorize publishing a rate.
+
+---
+
 ## [3.31.0] — 2026-09-05
 
 ### Added
@@ -20,24 +38,6 @@ All notable changes to this project are documented here.
 
 ---
 
-## [Unreleased] — 2026-09-04
-
-### Changed
-
-- **Published in the OpenAI Plugins Directory** as [Avoid AI Writing](https://chatgpt.com/plugins/plugins_6a9b77b18b8881918efa9c1255868164) (version 3.29.0, approved 2026-09-04). The bundled canonical skill now omits the frontmatter `metadata` block, which the portal rejects (#146); TERMS.md and PRIVACY.md state the plugin's scope and data handling in the terms OpenAI's plugin guidelines ask for (#147).
-- **Plugin validation now fails closed on deferred port-integrity gaps.** `agents/openai.yaml` rejects scalar policies and malformed mapping/list lines, SVG assets must have an actual `<svg>` root, and the bundled routing matrix carries a checked graph digest plus generated edge inventory so it cannot silently drift from `skill-graph.json`.
-
-- **README adds a restrained related-work block.** Links to Conor's public
-  builds, Chain of Thought, and `repo-audit` now sit after the core product and
-  usage documentation.
-- **Corpus manifest documents the register gap with two auditable seed entries.**
-  RFC 8259 provides a pre-LLM `docs` source and a 1995 W3C mailing-list message
-  provides a pre-LLM `conversational` source. The corpus README records that
-  both registers remain under-sampled and that `social` and `email` still have
-  no entries; the additions do not authorize publishing a rate.
-
----
-
 ## [3.29.0] — 2026-09-03
 
 ### Added