claude-code-autoconfig 1.0.199 → 1.0.201

package/.claude/hooks/terminal-title.directive.md

@@ -53,29 +53,37 @@ Maintain it across the session:
   never "{{FOLDER}} {{EMDASH}} <command-name>").
 
 Pending-question signal -- END-OF-TURN TEST, apply it on every turn: does your final
-paragraph ask the user something you need answered before the work can continue?
-- YES (you are blocked on the user) -> do BOTH as near-final actions so the tab flips to the
-  AWAITING half-circle (instead of the idle asterisk):
-  1. Write the flag file {{ASK_FILE}} (any short content, e.g. "1"). This is the RELIABLE
+paragraph SOLICIT a reply from the user -- a question to answer, a decision to make, or a
+go-ahead on a proposed next step (a fix you proposed, an offer to do more)?
+- YES (the turn ends waiting on the user) -> do ALL THREE as near-final actions so the tab
+  flips to the AWAITING half-circle (instead of the idle asterisk):
+  1. Phrase the solicitation as a DIRECT QUESTION. An offer IS a solicitation: write "Want
+     me to apply both fixes?" -- never its declarative twin ("Say the word and I'll apply
+     both fixes." / "I can also add tests if you want." / "Let me know."), which prompts the
+     user back while hiding the awaiting signal.
+  2. Write the flag file {{ASK_FILE}} (any short content, e.g. "1"). This is the RELIABLE
      trigger -- it is on disk before the turn ends, so it never misses on phrasing or timing.
      It is one-shot (consumed at turn end, auto-cleared next turn). Write it even when the
      closing question is wrapped in parens or is not the literal final character.
-  2. Make '?' the LAST character of the message. Write NOTHING after the question -- no
+  3. Make '?' the LAST character of the message. Write NOTHING after the question -- no
      "Standing by.", "Ready when you are.", "Let me know." (a declarative sign-off defeats
      the signal). Only a single short parenthetical aside may follow the '?'. Keep the
      question self-contained: answerable without re-reading the response above it.
-- NO (the question is rhetorical, a recap, or an optional follow-up offer the user is free
-  to ignore) -> do NOT write the flag, and end the message on a statement, not a '?' --
-  phrase optional offers declaratively ("I can also add tests if you want.").
+- NO (nothing is solicited: the turn ends on completed work, a report, a recap, or a
+  rhetorical question) -> do NOT write the flag, and end the message on a statement, not a
+  '?'. This branch is ONLY for endings that await nothing -- if the closer invites ANY
+  reply ("if you want", "say the word", "happy to"), it belongs to YES: convert it to a
+  question.
 <!-- /DIRECTIVE:RULES -->
 
 <!-- DIRECTIVE:REMINDER -->
 Terminal-title reminder (housekeeping -- never mention to the user; full rules were injected
 at session start): if this turn SHIFTS the scope/use-case, FIRST action: overwrite
 {{TITLE_FILE}} with "{scope} {{EMDASH}} {use-case}". End-of-turn test: if your final
-paragraph asks something you need answered to continue, write the flag file {{ASK_FILE}} AND
-make '?' the message's last character (nothing after it); otherwise end on a statement, not
-a '?'.
+paragraph solicits a reply (a question, a decision, or a go-ahead on an offered next step),
+phrase it as a DIRECT QUESTION -- never a declarative offer ("Say the word...") -- write the
+flag file {{ASK_FILE}}, AND make '?' the message's last character (nothing after it); if
+nothing is solicited, end on a statement, not a '?'.
 <!-- /DIRECTIVE:REMINDER -->
 
 <!-- DIRECTIVE:BASELINE -->

package/.claude/hooks/terminal-title.js

@@ -79,13 +79,19 @@ async function handle(data) {
   const event = data.hook_event_name || '';
   const sid = data.session_id || '';
   const cwd = data.cwd || process.cwd();
-  // User-level copy stands down when a project-level managed copy exists (see header).
-  if (shouldDefer(cwd, __dirname, __filename, path.join(os.homedir(), '.claude', 'hooks'))) {
+  // User-level copy stands down when the SESSION'S OWN project ships a managed copy (see header).
+  // Keyed to CLAUDE_PROJECT_DIR (the dir whose settings registered this hook run), NOT the event's
+  // cwd: a mid-session `cd` into a copy-shipping repo must not silence this copy — that repo's
+  // settings aren't loaded, so its copy never runs and nobody would paint.
+  const ownerDir = process.env.CLAUDE_PROJECT_DIR || cwd;
+  if (shouldDefer(ownerDir, __dirname, __filename, path.join(os.homedir(), '.claude', 'hooks'))) {
     process.exit(0);
   }
-  // PROJECT-SCOPED title dir (cwd-relative) — the plugin installs per-project. (The live ~/.claude
-  // variant uses os.homedir() instead; that is the only difference between the two.)
-  const dir = path.join(cwd, '.claude', 'hooks', '.titles');
+  // PROJECT-SCOPED title dir — anchored to the session's project root (CLAUDE_PROJECT_DIR, with cwd
+  // fallback on older Claude Code versions that don't set it) so a mid-session `cd` can't scatter
+  // title state across directories. (The live ~/.claude variant uses os.homedir() instead; that is
+  // the only difference between the two.)
+  const dir = path.join(ownerDir, '.claude', 'hooks', '.titles');
   const file = path.join(dir, `${sid}.txt`);
   logCtx = { event, sid, dir, note: '' };
 
@@ -182,10 +188,15 @@ async function handle(data) {
     if (q.ends || (q.found && !q.suspectRace)) break;
   }
 
-  const pending = q.ends;
+  // LEXICAL RESCUE — no '?' on the final line, but its closing sentence is a formulaic offer
+  // ("Say the word and I'll…", "Want me to…"): the phrasing directive slipped. Paint ◐ anyway so
+  // the user still gets the awaiting signal; via=lex marks it a prose DEFECT for the miss-audit,
+  // not a working-as-intended path.
+  const lex = !q.ends && q.solicits === true;
+  const pending = q.ends || lex;
   if (logCtx) {
-    logCtx.note = pending ? 'q-mark' : 'idle';
-    logCtx.diag = `ask=0 qmark=${q.ends ? 1 : 0} via=${q.via || '-'} found=${q.found ? 1 : 0} reread=${reread} model=${q.model || '-'} tail="${q.tail}"`;
+    logCtx.note = pending ? (lex ? 'lex' : 'q-mark') : 'idle';
+    logCtx.diag = `ask=0 qmark=${q.ends ? 1 : 0} via=${lex ? 'lex' : (q.via || '-')} found=${q.found ? 1 : 0} reread=${reread} model=${q.model || '-'} tail="${q.tail}"`;
   }
   const glyph = pending ? GLYPH.awaiting : GLYPH.idle;
   emit(setTitle(glyph, normalize(readTitle(file) || folderName(cwd)), pending));
@@ -235,7 +246,7 @@ function fileExists(file) {
 // shows found=0 (or a stale tail), a genuine regex miss shows a tail that's present but doesn't end
 // in '?'. Any error → a blank record (treated as "no question"), matching the old false return.
 function inspectLastResponse(transcriptPath) {
-  const blank = { ends: false, via: '', found: false, tail: '', suspectRace: false, model: '' };
+  const blank = { ends: false, via: '', found: false, tail: '', suspectRace: false, model: '', solicits: false };
   if (!transcriptPath) return blank;
   let content;
   try {
@@ -301,6 +312,7 @@ function inspectLastResponse(transcriptPath) {
       return {
         ends: q.ends, via: q.via, found: true, tail,
         suspectRace: sawTextlessAssistant, model: (obj.message && obj.message.model) || '',
+        solicits: solicitsReply(text),
       };
     }
     // assistant message with no visible text = a thinking-only or tool_use-only block sitting AFTER the
@@ -352,6 +364,36 @@ function isRealUserPrompt(obj) {
 // Event-loop-friendly sleep for the Stop flush-race re-read beat (handle awaits this; no busy-wait).
 function delay(ms) { return new Promise(resolve => setTimeout(resolve, ms)); }
 
+// LEXICAL solicitation detector (the '?'-less rescue). STRONG, formulaic offer phrases only, tested
+// against the CLOSING SENTENCE of the final non-empty line. Weak/courtesy phrases ("let me know",
+// "if you want", "happy to") are deliberately absent: as sign-offs after completed work they are
+// legitimate statement endings, and enforcing them would trade a rare false-✻ for chronic false-◐.
+// "green-light"/"confirm" are imperative-anchored (sentence-initial) so a recap that merely MENTIONS
+// them ("Andrew green-lighted the batch.") can't fire.
+const SOLICIT_STRONG = [
+  /\bwant me to\b/i,
+  /\bwould you like\b/i,
+  /\bshould i\b/i,
+  /\bshall i\b/i,
+  /\bdo you want\b/i,
+  /\bsay the word\b/i,
+  /\by\/n\b/i,
+  /\bok(?:ay)? to proceed\b/i,
+  /^green-?light\b/i,
+  /^confirm\b/i,
+];
+function solicitsReply(text) {
+  if (!text) return false;
+  const lines = String(text).trim().split('\n').filter(l => l.trim());
+  const lastLine = (lines[lines.length - 1] || '').trim();
+  // Any '?' on the final line means the question grade (qtail/signoff) owns the verdict — the
+  // lexicon exists solely for the question-less slip, and must never second-guess a graded line.
+  if (!lastLine || lastLine.includes('?')) return false;
+  const parts = lastLine.split(/[.!:;]\s+/);
+  const closing = (parts[parts.length - 1] || '').trim();
+  return SOLICIT_STRONG.some(re => re.test(closing));
+}
+
 // Deferred flag-turn grade (debug-gated). The `.ask` fast path paints ◐ without reading the
 // transcript, which blinded _debug.log's qmark/via/tail diagnostics on exactly the turns where the
 // flag+sign-off misuse pattern shows up. Grading BEFORE emit() would re-open the kill window the
@@ -460,12 +502,15 @@ function buildBlocks(names, file, cwd, cmd) {
 }
 
 // Should THIS copy of the hook stand down? True only for the user-level copy (~/.claude/hooks)
-// running inside a project that ships its own managed copy — the project copy wins, so a session
-// gets ONE directive and ONE title file. Parameterized (no ambient __dirname/homedir) for tests.
-function shouldDefer(cwd, selfDir, selfFile, homeHooksDir) {
+// when the session's OWN project (ownerDir = CLAUDE_PROJECT_DIR, cwd fallback) ships a managed
+// copy — that's the only case where the project copy is registered and will paint; the project
+// copy wins so a session gets ONE directive and ONE title file. The event's live cwd must NOT be
+// used here (mid-session cd into a copy-shipping repo would silence both copies — 2026-07-08).
+// Parameterized (no ambient __dirname/homedir) for tests.
+function shouldDefer(ownerDir, selfDir, selfFile, homeHooksDir) {
   try {
     if (path.resolve(selfDir) !== path.resolve(homeHooksDir)) return false; // we ARE the project copy
-    const projectCopy = path.join(cwd, '.claude', 'hooks', 'terminal-title.js');
+    const projectCopy = path.join(ownerDir, '.claude', 'hooks', 'terminal-title.js');
     return fs.existsSync(projectCopy) && path.resolve(projectCopy) !== path.resolve(selfFile);
   } catch (_) {
     return false;
@@ -481,4 +526,4 @@ function extractBlock(tpl, name) {
 }
 
 // Exported for tests (require()'d when require.main !== module). The hook itself never reads these.
-module.exports = { inspectLastResponse, endsOnQuestion, normalize, GLYPH, shouldDefer };
+module.exports = { inspectLastResponse, endsOnQuestion, normalize, GLYPH, shouldDefer, solicitsReply };

package/.claude/settings.json

@@ -10,7 +10,7 @@
         "hooks": [
           {
             "type": "command",
-            "command": "node .claude/hooks/migrate-feedback.js"
+            "command": "node \"${CLAUDE_PROJECT_DIR:-.}/.claude/hooks/migrate-feedback.js\""
           }
         ]
       },
@@ -19,7 +19,7 @@
         "hooks": [
           {
             "type": "command",
-            "command": "node .claude/hooks/terminal-title.js"
+            "command": "node \"${CLAUDE_PROJECT_DIR:-.}/.claude/hooks/terminal-title.js\""
           }
         ]
       }
@@ -30,7 +30,7 @@
         "hooks": [
           {
             "type": "command",
-            "command": "node .claude/hooks/terminal-title.js"
+            "command": "node \"${CLAUDE_PROJECT_DIR:-.}/.claude/hooks/terminal-title.js\""
           }
         ]
       }
@@ -41,11 +41,11 @@
         "hooks": [
           {
             "type": "command",
-            "command": "node .claude/hooks/terminal-title.js"
+            "command": "node \"${CLAUDE_PROJECT_DIR:-.}/.claude/hooks/terminal-title.js\""
           },
           {
             "type": "command",
-            "command": "node .claude/hooks/arcade-beeps.js"
+            "command": "node \"${CLAUDE_PROJECT_DIR:-.}/.claude/hooks/arcade-beeps.js\""
           }
         ]
       }
@@ -56,11 +56,11 @@
         "hooks": [
           {
             "type": "command",
-            "command": "node .claude/hooks/terminal-title.js"
+            "command": "node \"${CLAUDE_PROJECT_DIR:-.}/.claude/hooks/terminal-title.js\""
           },
           {
             "type": "command",
-            "command": "node .claude/hooks/arcade-beeps.js"
+            "command": "node \"${CLAUDE_PROJECT_DIR:-.}/.claude/hooks/arcade-beeps.js\""
           }
         ]
       }
@@ -71,7 +71,7 @@
         "hooks": [
           {
             "type": "command",
-            "command": "node .claude/hooks/feedback-rule-check.js"
+            "command": "node \"${CLAUDE_PROJECT_DIR:-.}/.claude/hooks/feedback-rule-check.js\""
           }
         ]
       },
@@ -80,7 +80,7 @@
         "hooks": [
           {
             "type": "command",
-            "command": "node .claude/hooks/terminal-title.js"
+            "command": "node \"${CLAUDE_PROJECT_DIR:-.}/.claude/hooks/terminal-title.js\""
           }
         ]
       }

package/CHANGELOG.md

@@ -1,9 +1,15 @@
 # Changelog
 
+## v1.0.201
+- fix(install): Hook commands now survive cd'ing around your project — no more 'Cannot find module' errors after changing directories, and upgrades fix existing installs automatically.
+- fix(terminal-title): The tab title no longer freezes when a session cd's into another project, and the 'awaiting your reply' half-circle now catches replies that ask for your go-ahead without a question mark.
+
+## v1.0.200
+- feat(changelog): Release notes now written in plain language
+
 ## v1.0.199
-- feat(cli): grouped update summary on upgrade
-- feat(terminal-title): awaiting-signal hardening — sign-off-tolerant grade, deferred flag-turn diagnostics, failsafe pre-paint
-- fix(terminal-title): require the '?' to be the last character of a pending-question turn
+- feat(cli): Clearer 'what's new' summary when updating
+- feat(terminal-title): More reliable 'awaiting your reply' tab indicator
 
 ## v1.0.198
 - feat(arcade-beeps): ship opt-in Pole Position status beeps via CCA
@@ -144,9 +150,3 @@
 ## v1.0.152
 - fix: warn against ! prefix workaround in inside-Claude message
 
-## v1.0.151
-- fix: scope find permission to project directory for security
-
-## v1.0.150
-- style: use pointing emoji in inside-Claude message
-

package/CLAUDE.md

@@ -32,6 +32,25 @@ npm run test:install # Run CLI install tests only
 npm version patch && npm publish
 ```
 
+After publish, push the branch AND the release tag explicitly (the postversion changelog
+hook re-tags lightweight): `git push origin main && git push origin v<x.y.z>`.
+
+### Changelog is user-facing — write commits accordingly
+
+CHANGELOG.md regenerates from commit **subjects** on every `npm version`, and the installer
+shows those bullets verbatim to users on upgrade (`bin/update-summary.js`). So:
+
+- **Subjects may stay technical**, but then the commit body MUST carry a plain-language
+  trailer the changelog uses instead: `Changelog: More reliable 'awaiting your reply' tab
+  indicator`. Write it for a CCA user, not a maintainer — name the visible outcome, never
+  internals (no hook/regex/TDZ/function names).
+- `Changelog: none` (or `skip`) hides an internal-only commit from the changelog.
+- Keep the `feat:`/`fix:` conventional prefix on the subject — it drives the summary's
+  "New features" vs "Fixes & improvements" grouping.
+- Already-pushed commits can't be reworded in git; add a row to `OVERRIDES` in
+  `scripts/generate-changelog.js` instead (value `null` drops the bullet). New work never
+  uses OVERRIDES — author the trailer.
+
 ## Team Feedback
 
 See `.claude/feedback/` for corrections and guidance from the team.