Release 2.0.2: fix the hook self-check message (#31)

.claude-plugin/marketplace.json

@@ -11,7 +11,7 @@
       "source": "./",
       "displayName": "Simple English",
       "description": "Write plain, layman-readable English with the discipline of ASD-STE100 Simplified Technical English: short sentences, one word one meaning, active voice, condition before command, terms defined at first use.",
-      "version": "2.0.1",
+      "version": "2.0.2",
       "author": {
         "name": "AminBlg"
       },

.claude-plugin/plugin.json

@@ -2,7 +2,7 @@
   "name": "simple-english",
   "displayName": "Simple English",
   "description": "Write plain, layman-readable English with the discipline of ASD-STE100 Simplified Technical English: short sentences, one word one meaning, active voice, condition before command, terms defined at first use.",
-  "version": "2.0.1",
+  "version": "2.0.2",
   "author": {
     "name": "AminBlg"
   },

.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "simple-english",
-  "version": "2.0.1",
+  "version": "2.0.2",
   "description": "Write plain, layman-readable English with the discipline of ASD-STE100 Simplified Technical English: short sentences, one word one meaning, active voice, condition before command, terms defined at first use.",
   "author": {
     "name": "AminBlg"

CHANGELOG.md

@@ -0,0 +1,50 @@
+# Changelog
+
+Each entry names the version, the date, and the measured effect where one exists.
+
+## 2.0.2, 2026-09-08
+
+- Fixed: the PostToolUse hook message said "Run the self-check in SKILL.md
+  before you deliver." At least one harness read that as an instruction to
+  search the install folder for a file. It read this instead of running the
+  check inside the skill already loaded. The message now names the change
+  directly. (#31)
+
+## 2.0.1, 2026-09-04 to 2026-09-06
+
+- Rebuilt the skill around two registers, the document and the reply, after
+  an audit found the old benchmark measured obedience to the linter, not what
+  a reader sees. SKILL.md is 1,825 tokens. The 53-rule catalog moved to
+  `references/rule-catalog.md`. Visible reply defects (over-cap sentences,
+  em-dashes, bold, headers, bullets) fell from 406 to 58 pooled over 16
+  replies on claude-sonnet-4-6. A blind judge preferred 2.0.1 over 2.0.0 in
+  14 of 16 pairs. Full method: `evals/results/rebuild-2026-09-02/RESULTS.md`.
+- Added `evals/check_numbers.py`, which recomputes every published number
+  from the raw files and fails the build on a mismatch. Wired into CI.
+- Fixed the Codex hook file collision: Claude Code auto-discovers
+  `hooks/hooks.json` at the plugin root, so the Codex configuration moved to
+  `.codex-plugin/hooks.json`.
+- Fixed: table rows in Markdown were blanked whole by the linter, which
+  exempted every cell inside them from every check. Each cell is now its own
+  sentence unit.
+- Fixed: the PostToolUse hook linted Claude's own memory and configuration
+  files. It now skips the Claude configuration directory and any path a
+  `SIMPLE_ENGLISH_LINT_EXCLUDE` glob names.
+- Fixed: the `claude-opus-4-8` benchmark row (1.05 baseline) did not
+  reproduce. Re-run with a clean baseline: 3.64. Old files moved to
+  `evals/results/raw-superseded/`.
+- Removed `references/checklist.md`. Nothing loaded it, and its content
+  restated the self-check, the catalog, and the linter.
+- Took the ASD-STE100 word-list extractor and the word-choice linter from a
+  closed pull request, without the extracted word lists themselves: the
+  standard forbids reproducing them without written authority from ASD. Users
+  build the lists locally from their own copy of the free PDF, in
+  `tools/ste-dictionary/`.
+- Added claude-fable-5-1 and claude-fable-5 rows to the linter benchmark, and
+  seven opencode model rows.
+
+## 2.0.0, 2026-09-01
+
+- Pivoted the default mode from strict ASD-STE100 enforcement to Plain: the
+  same structural rules, plus rules for readers outside the field and a
+  reply register, with Strict kept as a document-only overlay.

README.md

@@ -12,7 +12,7 @@
   <a href="evals/results/rebuild-2026-09-02/RESULTS.md"><img src="https://img.shields.io/badge/reply_defects-%E2%88%9286%25_measured-brightgreen?style=flat" alt="86% fewer visible reply defects, measured"></a>
   <a href="evals/results/RESULTS.md"><img src="https://img.shields.io/badge/benchmarked_on-9_Claude_models-blueviolet?style=flat" alt="9 models benchmarked"></a>
   <a href="https://agentskills.io"><img src="https://img.shields.io/badge/SKILL.md-open_standard-blue?style=flat" alt="Agent Skills"></a>
-  <a href="skills/simple-english/SKILL.md"><img src="https://img.shields.io/badge/version-2.0.1-blue?style=flat" alt="version 2.0.1"></a>
+  <a href="skills/simple-english/SKILL.md"><img src="https://img.shields.io/badge/version-2.0.2-blue?style=flat" alt="version 2.0.2"></a>
   <a href="LICENSE"><img src="https://img.shields.io/badge/license-MIT-lightgrey?style=flat" alt="MIT"></a>
   <a href="https://github.com/AminBlg/SimpleEnglish/stargazers"><img src="https://img.shields.io/github/stars/AminBlg/SimpleEnglish?style=flat&logo=github&color=yellow" alt="GitHub stars"></a>
 </p>

skills/simple-english/SKILL.md

@@ -17,7 +17,7 @@ description: |
 license: MIT
 compatibility: claude-code cursor codex gemini-cli opencode
 metadata:
-  version: "2.0.1"
+  version: "2.0.2"
   standard: ASD-STE100 Issue 9 (2025-01-15)
 ---
 

src/hooks/lint_hook.py

@@ -92,7 +92,7 @@ def post_tool_use(event):
     summary = ", ".join(f"{k} {v}" for k, v in hits.items())
     sys.stderr.write(
         f"simple-english: {target.name} has {report['violations_total']} STE violations "
-        f"({summary}). Run the self-check in SKILL.md before you deliver.\n"
+        f"({summary}). Fix these hits in the file you just wrote, then continue.\n"
     )
     return 2